No1 Rock-Bottom Spreads

New York
October 25, 2001
J.P. Morgan Securities Inc.

Portfolio Research
Peter Rappoport (1-212) 834-7046
rappoport_p@jpmorgan.com
Investment Strategies
www.morganmarkets.com
Rock-Bottom Spreads
Overview:
Fixed income investors have at their disposal an array of analytical tools for valuing
their interest rate exposure. However, most bonds present significant additional
exposures, notably credit and liquidity, for which valuation tools are scarce or nonexistent. This volume describes the rock-bottom spread framework we have developed as a first step to filling this void. Our aim is to make it possible to value the
credit exposure of fixed income instruments, much as their interest rate exposure
can be valued.
The cashflow pattern, maturity and seniority of the bond;

Its credit quality as determined by a rating;
Your views on broad credit trends over the bonds life; and
The rate of return you require in general for taking risk.
So, the bonds rock-bottom spread translates its promised cashflows, viewed from
your perspective on credit conditions, into the spread that will deliver you sufficient
return for bearing its credit risk. If its market spread falls short of rock-bottom,
you have a clear signal that holding the bond is not a good idea. It does not even
pay you enough for its credit exposure, never mind the liquidity you will forgo in
holding it. If the market spread exceeds rock-bottom, then the bond will be a good
buy if the excess spread is enough to compensate for its illiquidity. Figure 1 illustrates this basic valuation recipe.
Figure 1
Rock-Bottom Valuation: the basic recipe
Bond Characteristics,
Credit Rating
Credit Views
Risk/return target
Market Spread
below Rock-Bottom
Do not buy the

bond
Market Spread
above Rock-Bottom
Buy if surplus spread

pays for illiquidity
Rock-Bottom
Spread
INVESTMENT STRATEGIES: NO. 1
A bonds rock-bottom spread is what you, as an investor, need to be paid to bear its
credit exposure. At any lower spread, you will simply not earn enough to compensate for the credit risk the bond exposes you to, hence the term, rock-bottom. It
reflects four distinct components:
New York
October 25, 2001

Portfolio Research
So, rock-bottom spreads allow you to remove credit

from the valuation equation, much as duration and convexity allow you to remove interest rate exposure. Comparing the excess of market spreads over rock-bottom
across bonds or bond classes is frequently sufficient to
reveal significant valuation anomalies in market pricing.
For example
Figure 2
Rock-Bottom versus Market Spreads for US Corporate Bonds

Rock-bottom
exceeds
market
spread
Investment Strategies: No. 1

page 2
Figure 3
Anatomy of rock-bottom spread differences

EMBI Global with
current rating
outlooks
488bp
with stable
rating outlooks
514bp
+26bp
900
800
700
and with HY
credit composition
671bp
+157bp
and with historical

HY recovery rate
473bp
-198bp
and with historical

HY credit migration
691bp
+218bp
and with HY
diversity
491bp
-200bp
and with current HY

recovery rate view
551bp
+60bp
and with current HY

credit migration view
648bp
+97bp
600
Rock-bottom
less than
market
spread
500
400
300
Positive
liquidity
spread
200
100
AA
Rock-bottom
spread
BBB
BB
Market
spread
B-rated corporates have rarely paid in excess of their

rock-bottom spreads, throughout the bull and bear
markets of recent years. In contrast, over the same
period, BB-rated corporate market spreads have consistently exceeded rock-bottom by upwards of 100
basis points. Figure 2, taken from the article RockBottom Spread Mechanics in this volume, illustrates this
point with data from June 2001.
High Yield Corporate and Emerging Markets Sovereign
bonds have very different mixes of credit exposure,
making comparison of their raw market spreads a
treacherous indicator of relative value. However, the
rock-bottom spread framework allows us to price each
of these differences separately. This provides a framework for identifying and taking views on the key
differences between the two markets. For example, the
decomposition of the two markets rock-bottom
spreads in Figure 3, versions of which are used in the
Emerging Markets versus High Yield articles, points to
differences in (assumed) recovery rates as the single
largest contributor to the rock-bottom spread differential.
Over the last few years, a strategy that sold bonds with
higher rock-bottom than market spreads, and bought
HY with current
default view
648bp
those with lower rock-bottom spreads than market,

outperformed the High Grade market by about 1.1%
annually, and outperformed the Speculative Grade
market by about 6% annually. Figures 4 and 5 document that this outperformance has occurred consistently, and with low downside.
This collection of articles is designed to enable you to
use the rock-bottom spread framework in your credit
investment decisions, whether strategic or tactical. It
draws together articles JPMorgan has written and
resources it has developed to calculate rock-bottom
spreads, understand what drives the calculations, and
explore where they can add value. Our review is organised in three sections.

page 3
1. Understanding Rock-Bottom Spreads
Figure 4
The first paper in this volume, Valuing Credit Fundamentals: Rock-Bottom Spreads, develops the motivation for rock-bottom spreads in detail. As with any
analytical tool, it is necessary to feel comfortable with
how rock-bottom spreads work, and to understand why
the calculations deliver the results they do. To this end,
we have included Rock-Bottom Spread Mechanics
which shows how rock-bottom spreads follow logically
once you have set yourself a target risk-adjusted return,
or information ratio. It then shows how to calculate
rock-bottom spreads, starting from the simplest example of a one-year bond, and moving on to longermaturity bonds. For each input to the rock-bottom
spread calculation, such as a bonds coupon, maturity
and assumed recovery rate, it shows how the resulting
rock-bottom spread changes as the input changes. This
gives an idea of the sensitivity of rock-bottom spreads
to changes in bond characteristics and assumptions, as
well as a sense of which drivers of the rock-bottom
spread are important. All of the examples can be replicated using the rock-bottom spread calculator on our
MorganMarkets website, by following the recipes in the
paper.
Annualized return relative to Index
Another resource for understanding rock-bottom

spreads, Rock-Bottom SpreadSheet, is available on
MorganMarkets. This interactive presentation builds up
a complete calculator step-by-step, making it possible to
explain the source of each number in each step. This
allows you to concentrate on particular steps of the
calculation, in a way that would not be possible in a
research paper.
2. The Rock-Bottom Spread Web Calculator

Calculating rock-bottom spreads is a purely mechanical
matter, once you have assembled the appropriate
inputs bond characteristics, market views, etc. The
principal problem is to manage and manipulate the large
3%
2%
-1%
Average
2001I
2000II
1999II
1999I
1998II
1998I
0%
1997II
1%
Figure 5
US High Yield Rock-Bottom Bondpicking Strategy

Annualized return relative to Index
12%
10%
8%
6%
4%
TABLE
OF
Average
2001.I
0%
2000.I
2%
1999.II
One of the principal drivers of rock-bottom spreads is

the pattern of credit migration, that is, the frequencies
with which upgrades, downgrades and defaults occur.
The most complete information is available for US
issuers, but there is a widespread view that different
credit migration assumptions are relevant for European
issuers. In Valuing European Credit Fundamentals,
we assess the implications for credit spreads of the
different credit rating experience of European issuers.
US Investment Grade Rock-Bottom Bondpicking Strategy
2000I

Portfolio Research
2000.II
New York
October 25, 2001
CONTENTS:
1. Understanding Rock-Bottom Spreads

Valuing Credit Fundamentals: Rock-Bottom Spreads
Rock-Bottom Spread Mechanics
17
Valuing European Credit Fundamentals
29
2. The Rock-Bottom Spread Web Calculator

An Annotated Introduction to the Rock-bottom Calculator
35
Introducing the Rock-Bottom Roundup
43
Rock-Bottom Roundup
45
3. Rock-Bottom Investment Strategies

Picking High Yield Bonds
49
Picking Investment Grade Bonds
57
Valuing Rating-triggered Step-up Bonds
67
Comparing Credit Fundamentals:

Emerging Markets versus High Yield
69
Emerging Markets versus High Yield:

Credit Fundamentals Revisited
77
US Credits Look Attractive for Japanese Investors
89
New York
October 25, 2001

Portfolio Research
amount of data involved. This problem is solved by the

rock-bottom spread calculator, which makes it simple to
take views on credit trends, calculate the resulting rockbottom spreads on any portfolio of bonds, and examine
the sensitivity of these spreads to changes in your views.
The calculator is to be found on our MorganMarkets
website [www.morganmarkets.com], and is described in
Using the JPMorgan Rock-Bottom Spread Calculators.
The calculator makes it possible to trawl through literally
thousands of bonds, and value them all in a common
framework. You can either upload your own portfolio
details, or examine the results for groups of bonds that
are permanently accessible to the calculator, such as our
High-Yield Corporate and Emerging Markets sovereign
indices, and High-Grade Corporates.
Since many investors may find it useful to have information on these broad asset classes, a periodic publication
will summarize their rock-bottom and market spreads
under a basic credit view. A sample of this RockBottom Roundup is included in this volume, along with
an explanation of how to read it. The Rock-Bottom
Roundup compares market and rock-bottom spreads
across the three asset classes, as well as breaking them
down by rating category and seniority.
3. Rock-Bottom Investment Strategies

A bond whose market spread exceeds its rock-bottom
spread is cheap, in the sense that its expected future
returns are high relative to the anticipated credit risk.
Picking High Yield Bonds and Picking Investment
Grade Bonds develop this idea, and show that bond
selection rules based on it have substantially outperformed the High Yield and Investment Grade corporate
markets since 1997. In each case, we examine in depth
the source of the rules outperformance, and its limitations. In both cases, it is quite clear that the rock-bottom
rule identifies a systematic source of value that is not
picked up by other, more traditional approaches to bond
selection. In particular, using the rock-bottom framework is the obvious way to apply the same yardstick to
all bonds a highly desirable feature for any relative
value discipline.

page 4
One of the virtues of the rock-bottom framework is that

it enables you to place a value on highly complex
streams of contingent cashflows, as shown in Valuing
Rating-triggered Step-up Bonds, which focuses on
recently issued telecom paper. One of the features to
which it draws attention is the different value of alternative step-up schedules.
We also include two earlier pieces of research, Comparing Credit Fundamentals: Emerging Markets versus
High Yield, and Emerging Markets versus High
Yield: Credit Fundamentals Revisited. These papers
detail in particular the relative diversity of Emerging
Markets Sovereigns and High Yield Corporates, and the
credit migration assumptions we use for Emerging
Markets in the Rock-Bottom Spread Calculator.
Last, US Credits Look Attractive for Japanese Investors evaluates opportunities for Japanese investors in
the US corporates, by calculating the surplus spread
they offer over rock-bottom in Yen terms, and comparing it with the surplus offered by Japanese issuers.
Since the articles in this collection were originally issued
as standalone pieces we do not recommend reading
from cover to cover. Rather, this volume is best used as
a reference.
Each article contains a description of how the RockBottom Spread valuation framework operates, and so,
we hope, will be understandable to anyone not versed in
Rock-Bottom Spreads as an application of the concept.
For a detailed understanding of Rock-Bottom Spreads,
we recommend that you read the first article, Valuing
Credit Fundamentals, to provide an overview, but
concentrate on the examples in Rock-Bottom Spread
Mechanics to gain a thorough grounding in what drives
rock-bottom spreads. The examples in this article are
designed to be easily replicable on our web calculator,
and we recommend that the two be used together.
New York
October 25, 2001

Portfolio Research

page 5
Originally published on
November 17, 1999
Portfolio Research
Valuing Credit Fundamentals:

Rock-Bottom Spreads
Rock-bottom spreads measure the lowest acceptable reward
for bearing exposure to default and downgrades
Comparing rock-bottom spreads to market spreads is
central to assessing the value of credit instruments
High-grade market spreads greatly exceed rock-bottom
spreads, while speculative grade spreads often fall short of
rock bottom.
- this means that market spreads pay the lowest liquidity
premium for the least liquid bonds
This years deterioration in credit fundamentals
- is priced into high-grade spreads
- is overcompensated by BB new issue spreads
- is not reflected by other high-yield spreads
Do BB-rated spreads currently offer better or worse value than
investment grade spreads? Is the rise in A-rated spreads this
year sufficient to compensate for growing worries about credit?
Any informed view on these questions needs to take account of
credits many moving parts: default rate trends, recovery rates,
changes in credit quality. Not making these credit fundamentals
explicit runs the risk of losing them in the shuffle.
This note shows how to value a bonds credit fundamentals
exposure, which we translate into its rock-bottom credit spread.
This is the lowest spread at which an investor should be
willing to bear its credit fundamentals exposure.
The difference between market spreads and rock-bottom
spreads is, accordingly, the maximum available to compensate
for corporates lower liquidity than governments. We refer to
this as the illiquidity spread. Decomposing spreads into credit
fundamentals and liquidity components provides a new and
surprising perspective on the sources of value in credit markets.
Investment and speculative grades offer very different
packages of payment for illiquidity and credit fundamentals
(Figure 1). Investment grade rock bottom spreads are very low
relative to market spreads. AA-rock-bottom spreads of 18bp
compare to market AA spreads of 90bp. This leaves 72bp as
compensation for AAs greater illiquidity than governments.
Moving down the credit spectrum, rock-bottom spreads
increase, as we would expect. In fact, they grow faster than
Figure 1
Rock-Bottom Spreads vs. Market Spreads,

October 1999 (10-year maturity)
Basis points
800
700
Rock-Bottom
Spread
600
500
400
300
Market Spread
200
Illiquidity
Spread
100
0
-100
-200
Spreads
Market
Rock-Bottom
Illiquidity
AAA
AA
72
6
66
90
18
72
124
34
90
BBB
BB
165
99
66
300
296
4
510
713
-203
Rock-bottom spreads calculated from the credit downswing scenario

described in Section 5, below
market spreads. Speculative grade market spreads are

dominated by rock-bottom spreads, with B-rated spreads
even falling short of rock-bottom by 203bp. High-grade
illiquidity spreads exceed high yield (Figure 1), even
though speculative grades are much less liquid. This
means that the market is not pricing credit fundamentals
and liquidity exposures consistently across the credit
quality spectrum.
Not only are investment grade rock-bottom spreads low
relative to market spreads, they also move little when credit
fundamentals change. Figure 2 compares rock-bottom
spreads arising from two credit fundamentals scenarios,
one conforming to historical average default rates, the
other a credit downturn scenario, in which adverse credit
conditions are expected over the next two years. AA rock
bottom spreads differ by just 2bp. High-grade investors
who focus on managing their credit fundamentals
exposure at the expense of their liquidity exposure, would
seem to be misallocating their efforts.
In contrast, as credit fundamentals change, significant
movements in speculative grade rock-bottom spreads
occur (Figure 2). Assessing whether changes in credit
fundamentals are matched by market spreads is essential
New York
October 25, 2001

Portfolio Research
Figure 2
Adverse Credit Outlook: Change in Rock-Bottom Spreads

Basis points
120

page 6
6. Interpreting Market Spreads ....................................... p.8

We use the credit fundamentals views contrasted in Figure 2
above to cast light on how the market has responded to the
deterioration in credit conditions this year.
100
A rock-bottom spread calculator is available on the Credit

page of our website, Morgan Markets, and is explained in a
separate document, available on the website.
80
60
40
20
0
AAA
Roc k-Bottom S p rea d s
Credit D o wn turn
Sce nar io
6
Histo ric al Av erage
Sce nar io
6
Dif ferenc e
0
AA
BBB
BB
18
34
99
296
7 13
16
2
29
5
84
15
259
37
6 13
1 00
for high-yield investors, and is made easier by tracking rockbottom spreads.

The body of this paper is organized into these sections:
1. Credit Fundamentals .....................................................p.3
We describe how historical default rates, recovery rates, and
changes in credit quality affect rock-bottom spreads.
2. Credit Returns ..............................................................p.4
We describe the precise features of corporate bonds priced
by rock-bottom spreads.
3. Risk Tolerance ..............................................................p.4
We describe risk tolerance in terms of a target information ratio
that credit returns need to attain, in order to be competitive
with other alternatives to investment in government bonds.
4. Rock-Bottom Spreads ....................................................p.5
We describe how we calculate rock-bottom spreads from
credit fundamentals information and the investors risk
tolerance, measured by a target information ratio. This
section may be cheerfully skipped by anyone not interested
in the mechanics of the calculations.
5. Forecasting Credit Fundamentals .................................p.6
Rock-bottom spreads are essentially forward-looking, and the
credit fundamentals on which they depend amount to a
forecast of the future course of credit conditions. Credit
fundamentals have many moving parts, and forecasting each
is simply not practical. We show how to boil credit
fundamentals down to a very small number of features on
which a view needs to be held.

Portfolio Research
1. Credit Fundamentals
Ultimately, credit fundamentals concern the potential for
losses from default (Figure 3). Thus, a bonds short-term
likelihood of a default and how much will be recovered in
default both figure. Similarly, the bonds credit quality may
change, altering its subsequent default probability. Losses
will also be easier to bear, the greater the diversity of the
portfolio in which the bond is held.
To arrive at rock-bottom spreads, we need to quantify each
of these credit fundamentals elements. As a baseline
scenario, we use historical averages for the default profile.
We use the transition probabilities tabulated by Moodys
to measure the likelihood of each possible change in credit
quality*. The achievable diversity is measured from recent
estimates of the size of the corporate bond market (Figure 4).
A few concrete examples best illustrate how these numbers
drive a bonds credit fundamentals exposure. The top right
number in the transition matrix in Figure 4 indicates that
AAAs have zero probability of default over one year.
Consequently, a one-year AAA has zero credit fundamentals
exposure, and its rock-bottom spread will also be zero. The
bonds 10.3% chance of being downgraded to AA, 1% chance
of a downgrade to A, and so on, are irrelevant for a one-year
bond, because it will still pay in full even if it is downgraded.
However, if the bond has more than a year to go, these
positive probabilities of downgrade during its first year open
the possibility that it can default in its second year. For
example, if the bond is downgraded to A after one year, it then
has a 0.01% chance of default in its second year. Of course,
both the probabilities involved in this 2-year AAA example
are minute (1% and .01%), and so contribute little to credit
fundamentals exposure. This, in turn, will translate into a very
small rock-bottom spread. Similarly, we would expect the
credit fundamentals exposure and spread of a B-rated bond to
be much higher, because with a one-year probability of 7.1%,
the event of default is far less remote.
Historically, recoveries from default have borne no relation to
the rating of the bond prior to default. Consequently, we
assume that any defaulting issue will enter a recovery process
which produces the historical average of $47 per $100 of
principal, albeit after 2.1 years on average. There is some
uncertainty about the actual recovery amount, captured by
the $26 volatility of recoveries.
* Standard and Poor provide similar tables. Note that we are equating the rating
categories of the two agencies (for example, Baa and BBB), and using Standard and
Poors nomenclature. These historical transition probabilities record the
frequency with which the agencies change ratings. We are thus assuming that
these frequencies are a reliable representation of how actual credit quality
changes. This does not require that the rating agencies anticipate changes in
credit quality, just that, ex post, the frequency of rating changes matches the
frequency of credit quality movements.

page 7
Figure 3
Components of Credit Fundamentals

Probability of
Credit Quality
Change
Portfolio
Diversity
Probability
of Default
Recovery Rate
Credit Fundamentals
Exposure
Figure 4
Credit Fundamentals: Historical Average Scenario

Probability of Credit Quality Change and Default (% per year):
Rating at year-end
AAA AA
Rating at start of year
New York
October 25, 2001
BBB
BB
CCC Default
AAA 88.7
AA
1.1
10.3
88.7
1.0
9.6
0.00
0.3
0.03
0.15
0.00
0.15
0.00
0.00
0.06
2.9
90.2
5.9
0.7
0.18
0.01
0.01
BBB
BB
0.05
0.03
0.3
0.08
7.1
0.6
85.2
5.7
6.1
83.6
1.0
8.1
0.08
0.5
0.16
1.5
0.01
0.04
0.2
0.7
6.6
82.7
2.8
7.1
CCC 0.00
0.00
0.7
1.1
3.1
6.1
63.0
26.2
0.00
0.01
Source: Moodys. The figures are long run annual averages of the
frequency of rating changes and defaults among rated issuers, 1980-98
Recovery Rate:
Average Recovery Rate:
$47 per $100 of principal
Volatility:
$26 per $100 of principal
Average time to recovery:

2.1 years
Source: H.S. Wagner The Pricing of Bonds in Bankruptcy and Financial
Restructuring, Journal of Fixed Income, June 1996
Portfolio Diversity:
Diversity Scores
AAA AA
BBB BB
CCC
30
66
63
54
19
53
59
The diversity score is the number of uncorrelated exposures to which each

sector is equivalent. See Appendix 1.
In all cases, an investor will be more prepared to take exposure

to credit fundamentals at a given spread, the more it is
possible to diversify the default risks. The extent of
diversification is captured by the diversity score in Figure 4.
Introduced by Moodys, this translates a portfolio of
correlated exposures into a smaller number of uncorrelated
exposures. (Diversity scores are further discussed in
Appendix 1). We calculate diversity scores as if the
investors portfolio were the entire U.S. corporate market.
Any less diversified portfolio would occasion a higher rockbottom spread than we calculate.
New York
October 25, 2001

Portfolio Research

page 8
2. Credit Returns
3. Risk Tolerance
The essence of credit risk is that promises to pay cashflows

may not be kept. Accordingly, it makes sense to compare
bonds exposed to credit risk to government bonds, which, for
our purposes, are those whose cashflow promises are not in
doubt.
Rock-bottom spreads are the result of combining credit

fundamentals with the investors risk appetite (Figure 6).
This motivates examining a bonds credit return, which is the

excess of its return over that of an identical bond whose
cashflows are promised by the government. To make our
exposition of rock-bottom spreads clearer, it is simplest to think
of the corporate bond and the government bond as trading at
par, and promising different coupons on the same date. The
corporate bond promises more, in the form of a spread over
government coupons, to compensate for the uncertainty of its
promises.*
Figure 5
Cashflow Patterns
Corporate Bond
Now pay...
1 year
At maturity
receive ...
Credit
Return
5% Government Bond
$100
Default
Recovery
$100
Figure 6
Rock-Bottom Spreads
Credit Fundamentals
Exposure
Risk
Tolerance
Rock Bottom
Spreads
The information ratio

We think of the investors neutral position as being investment in
government bonds that match liabilities cashflow-for-cashflow.
This is not only relevant to fund managers who are mandated to
outperform a government bond index. It is also a sensible baseline
for assets invested to meet defined benefit pension liabilities or life
insurance annuity contracts. If the liabilities are cashflow-matched
with government paper, there is no chance of failing to meet
liabilities. Credit exposure creates the possibility that liabilities will
not be met due to defaults, so an excess return over governments
will be demanded.
No default
$100 + $5
+
Spread
Corporate
Return
$100 + $5
Government
Return
Figure 5 illustrates for the case of a one-year bond.

There are only two possible scenarios default and no
default (because the investment time horizon and maturity of
the bond coincide). The two bonds promised cashflows,
which are realized in the absence of default, differ only by the
spread paid on the coupon date. Their actual cashflows will
also differ when the credit risky bond defaults.
This simple example illustrates how credit fundamentals
exposures -- the possibility of a default, and the dimensions of
what is recovered in the event of a default, give rise to a credit
spread. The obvious question is, how much spread? For this,
we need to stipulate the risk tolerance of the investor.
We want to establish the spread at which it is worth holding a bond

involving credit fundamentals in preference to a government bond
with identical cashflows. To do so, we work backwards from the
general conditions that should induce an investor to hold any asset
in preference to government bonds, namely that the extra return
earned on average is sufficient compensation for the extra risks.
Here, the extra return that concerns us is the credit return.
We quantify this condition by requiring that the credit bond must
be priced to offer an excess average return per unit of excess return
volatility or information ratio commensurate with other
opportunities for outperforming governments. A rule-of-thumb
used by plan sponsors is that candidate new asset classes need to
demonstrate an information ratio of one-half. The historical
performance of active managers of global government bond
portfolios a competing use of the funds that might otherwise be
invested in credit is also one-half (see Maintaining returns in a
low-yield world, J.P. Morgan, January 7, 1999). Accordingly, we
require credit bonds to produce an information ratio of one-half. An
asset class with an information ratio of 0.5 will underperform its
benchmark (here governments) one year in every three.
The actual information ratio offered by a bond portfolio depends
only on:
the credit fundamentals features depicted in Figure 3,
the spread of the bonds.
* An alternative is to assign the government and corporate bonds the same coupon.
In this case, they do make identical promises, and the government bond will cost
more than the corporate as a consequence. The two perspectives lead to
practically identical rock-bottom spread figures.
We can use this relationship to back out the spread that, in

combination with the bonds credit fundamentals, generates the
target information ratio of one-half. This will be the rock-bottom
spread.
New York
October 25, 2001

Portfolio Research
4. Rock-Bottom Spreads
We need to flesh out the relationship between spreads and credit
returns. We start with the one-year bond described in Figure 5,
whose current cost is $100, and whose annual coupon we think of
as equal to the government rate plus a spread. With an
investment horizon that is also one year, what is the minimum
spread over the one-year government rate at which you would
hold this bond, that is, the spread that causes the bonds
information ratio to be one-half?

page 9
Figure 7
Calculating One-Year BB-rated Rock-Bottom Spreads

1.5
(a) Average Credit Return (%)
1.0
0.5
Spread (bps)
0
50
-0.5
100
150
200
150
200
-1.0
-1.5
The information ratio is the ratio of average credit returns to the

volatility of credit returns. We now trace the relationship of each
to the bonds promised spread.
Average credit returns
The average credit return combines the returns in the two
scenarios, according to their probability of occurrence. We
illustrate with a BB-rated bond, so the relevant default rate is 1.5%
per year. Figure 7(a) shows that the average credit return
increases with the promised spread. The position of this
relationship is driven principally by the default probability. The
line shifts to the right as the default probability rises, so the Brated line would lie to the right of the BB line, while the A-rated
line would lie to the left.
If we were only concerned with breaking even with governments,
Figure 7(a) would be as far as we need go, since we could read off
the required spread level as the intersection of the average credit
return line with the horizontal axis. However, this 97bp does not
compensate at all for the risk of credit fundamentals.
Volatility of credit returns
Figure 7(b) charts the volatility of the credit returns of BB bonds
against promised spreads. It is important to keep in mind that the
volatility of credit returns is not the same as the volatility of
excess returns of spread instruments that we observe in the
market, since this contains the volatility of the liquidity
component as well. Instead, the volatility of credit returns is
effectively the range of variation between the $100 of principal
received in the absence of default, and the average of $47,
discounted back 2.1 years, that is paid out in default.
Consequently, volatility varies very little as the level of promised
spreads moves over a range of a few hundred basis points. (The
line in Figure 7(b) is very flat). However, the volatility line is
shifted upwards as the probability of default increases, and as
portfolio diversification declines.
Finding the rock-bottom spread
It now remains to divide the average credit return at each spread
by the corresponding volatility, to arrive at the actual information
ratio corresponding to that spread (Figure 7(c)). From this, we
can read off the rock bottom spread of 152bp, at which an
information ratio of one-half results. Comparing this figure with
the breakeven spread of 97bp reveals that our assumed risk
appetite demands a risk premium of 55bp for holding BB-rated
paper for one year.
Divided by...
1.2
1.0
0.8
(b) Credit Return Volatility (%)
0.6
0
50
100
equals...
1.0
(c) I nform ation Ratio

0.5
Target
Rock-Bottom
Spread: 152bp
0
50
100
150
200
Breakeven
Spread: 97bp
-0.5
Explaining Figure 7
First, some abbreviations: p denotes the probability of default
during the year, d denotes the excess return over governments in
the event of a default (recovery-100-coupon) and s denotes the
spread, n denotes the diversity score, or equivalent num ber of
independent credit exposures in the portfolio (S ee A ppendix 1).
The average credit return is the average of the exces s return
over governm ents in F igure 6, weighted by their probabilities:
p*d + (1-p)*s
Hence, average credit returns increase as s preads increase. As
the probability of default increases, the line flattens and shifts
dow nwards.
The volatility of returns is:
p * (1 p )
* (s d )
So volatility increases with spread (although not

significantly, s ince(s-d) is on the order of $68, and an extra
100bp of spread raises this by only $1). As divers ity
declines or the probability of default increases (up to one
half), the volatility line shifts up.
New York
October 25, 2001

Portfolio Research
These calculations can be carried out for any rating

category, as shown in Table 1. For lower credit qualities,
the probability of default rises, the average credit return
associated with any given spread level falls, and the
volatility rises. Thus, the spread necessary to deliver the
target information ratio also rises. Spreads for AAA-rated
bonds are zero, because there is zero probability of default
over one year, according to historical experience.
Table 1
One-year credit bonds

RockAnnual Default
Breakeven
bottom
Risk
probability (%)
spread (bp)
spread (bp)
premium (bp)
AAA
0.000
AA
0.005
0.010
BBB
0.16
10
28
18
BB
1.46
97
152
55
7.06
489
632
143
26.16
2282
2959
677
CCC
Rock-bottom spreads for longer maturities

This analysis is simple to extend to bonds with longer than
one year until maturity. What we needed to know to
calculate the one-year spread was the value of our bond at
maturity (par or default). To arrive at two-year rock-bottom
spreads, we need to know the value of the bond in each
credit fundamentals scenario (i.e., each rating category) at
the end of one year. But we have this information from the
one-year spreads in Table 1 (which readily translate into
one-year prices). This inductive calculation can be repeated
for each credit quality at each maturity, tracing out the entire
structure of rock bottom credit spreads by term to maturity
and credit quality (see Appendix 2). Selected maturities are
shown in Table 2. The 10-year spreads are those that appear
in Figure 2*, under the historical average scenario.
Table 2
Rock-Bottom Spread Term Structures

Basis points
Maturity
3
10
AAA
AA
16
15
29
28
42
56
84
BB
152
191
220
259
632
645
642
613
2959
2558
2239
1759
BBB
CCC
* Our spread calculations omit compensation for individuals extra state income tax
liability on corporate bonds, compared to US Treasuries. While rates vary from
zero to 12%, we have no estimates of the tax rate of the marginal investor, which is
the relevant rate. Consequently, we think of the residual illiquidity spread as
including compensation for differential taxation. While this complicates using
the illiquidity spread to measure total compensation for liquidity factor, it does
not affect comparisons of spreads between rating categories, since all are taxed the
same.

page 10
Some perspective
It is worthwhile at this point to take stock. First, the
framework we have used to arrive at rock-bottom spreads
may seem similar to that used to price options or other
derivatives. There are indeed great similarities. In particular,
we are using the tree type of structure to capture all the
possible scenarios. Here, the branches in the tree arise from
changes in ratings. However, the use to which we put this
framework is entirely different from the derivatives case. To
price an option, one takes the value of the underlying
security or securities as given, and ensures that the option is
priced so that no profitable arbitrage trade involving the
underlying security and the option is possible. In the case
of rock-bottom spreads, we are valuing not a derivative, but
the underlying security itself. There is therefore no riskless
portfolio (composed of the derivative and the underlying)
that we can set up that enables us to ignore both the risk
preferences of the investor, and the average return of the
securities. Instead, the risk preferences are described by the
investors target information ratio, and the average (credit)
return is the one that results from satisfying these risk
preferences in the context of the bonds credit fundamentals.
We thus do not price credit risk in a way that is consistent
with market prices (i.e., arbitrage-free), but rather provide a
valuation that is independent of the markets.
Second, rock-bottom spreads essentially reflect the
investors reservation spread for credit exposure. In
economics terminology, we have identified a point on the
investors demand curve. This is somewhat different from
other approaches to valuation of securities, for example, that
used in our Global Markets Outlook and Strategy (GMOS)
publication. The GMOS valuation framework makes
statements about where market yields are likely to move,
based on historical estimates of risk premia, and the
observed rate of mean reversion of yields to their equilibrium
values. Thus, both supply and demand sides of the market,
as well as adjustment to equilibrium, are brought under the
valuation umbrella. In the case of rock-bottom spreads, the
supply side of the corporate market is not addressed, and
there is no presumption that spreads will revert to a
equilibrium fair-value levels indicated by rock-bottom
spreads. In short, using rock-bottom spreads to value does
not require any belief that the market is also using them.
Instead, they reflect what the investor needs to be paid to
bear the credit fundamentals risks involved, which will be
dictated in part by the particular situation of the investor.
Here, we have chosen the situation of a quite representative
investor, one whose investment problem is to, or reduces to
attempting to, outperform a government benchmark. This
type of investor has the assurance that, if his/her credit
fundamentals views are correct, bond portfolios that yield in
excess of rock-bottom will produce a superior risk-return
performance than their target information ratio.
New York
October 25, 2001

Portfolio Research
5. Forecasting credit fundamentals

Rock-bottom spreads are essentially forward-looking. Thus,
Figure 4s historical average credit fundamentals scenario,
which we have used to illustrate how to calculate rockbottom spreads, are relevant only if one believes that the
historical average scenario will describe each year in the
future. However, in general, the credit quality environment
changes over time, and rock-bottom spreads should reflect
anticipated changes in credit fundamentals. Interpreted
literally, this means taking a view on each element of the
transition matrix (including default rates) for each year in the
future, which is obviously not practical. It is therefore
important to boil the necessary information down to a small
number of indicators, about which views can be formed.

page 11
The rating agencies regularly publish statistics on ratings

drift (number of upgrades minus downgrades), activity
(sum of upgrades and downgrades), and speculative grade
default rates. Figure 8 shows that these have indeed
fluctuated substantially around their long-term averages.
They seem to be sensible measures about which to form
views about future credit fundamentals. Rating drift is
analogous to an average change of credit quality, while rating
activity is analogous to volatility. Speculative grade defaults
track an important source of credit returns. It is a lot more
easy and intuitive to extend the lines in Figure 8 along an
expected future path, than it is to fill numbers in a transition
matrix. The challenge is to propagate these paths
appropriately into anticipated transition matrices.
Fortunately, a few simple rules appear to do the trick. These
are described in Appendix 3.
To illustrate the process, assume we forecast the credit
downswing shown in Table 3:
Figure 8
Aggregate Measures of Credit Quality

Annual percentage rates
Table 3
40
Credit Fundamentals Summaries
Rating activity
30
Credit Fundamentals Scenario

Historical
Credit
20
Average
Speculative Grade
Defaults
10
Average 20%
4.10%
6%
20%
20%
-8.50%
-12%
Activity
0
Dec 82
Dec 86
Dec 90
Dec 94
Dec 98
10
5
Which implies
Upgrades
5.75%
4%
Downgrades
14.25%
16%
The changes implied by the resulting transition matrix are

shown in Figure 9. There is a tilt from upgrades to
downgrades, and from live credit ratings to default, as
Rating drift
Average -8.5%
-5
Drift
Downswing
Figure 9
-10
Credit Downswing Scenario: Changes in Transition

Probabilities from Historical Average
-15
-20
Rating at year-end
Rating at start of year
-25
Speculative Grade Defaults

10
8
6
4
2
Average 4.1%
0
Dec 82
Source: Moodys
Dec 86
Dec 90
Dec 94
Dec 98
AAA
AA
A
BBB
BB
B
CCC
AAA
AA
-0.7
-0.4
+0.0
-0.0
-0.0
-0.0
+0.0
+0.6
-0.9
+0.0
-0.1
-0.0
-0.0
+0.0
+0.1
+1.3
-1.8
-2.5
-0.3
-0.1
-0.3
BBB
+0.0
+0.0
+1.5
+1.0
-3.1
-0.3
-0.5
BB
+0.0
+0.0
+0.2
+1.3
+0.1
-3.4
-1.4
B
+0.0
+0.0
+0.0
+0.2
+2.5
-0.7
-2.8
CCC
+0.0
+0.0
+0.0
+0.0
+0.2
+2.0
-2.0
D
+0.0
+0.0
+0.0
+0.1
+0.7
+2.5
+7.0
would be expected. This transition matrix is used to generate

the rock-bottom spreads for the credit downturn scenario
discussed in the Introduction. We assume that it persists for
the next two years, after which credit fundamentals revert to
the historical average transition matrix shown in Figure 4.
New York
October 25, 2001

Portfolio Research

page 12
6. Interpreting Market Spreads
Figure 1
The object of calculating rock-bottom spreads is to provide

perspective on market spreads.
Basis points
Serious concern over adverse credit conditions has only

emerged during the middle of the year, as speculative
defaults rose, and downgrades accumulated and once again
outstripped upgrades. Earlier in the year, credit
fundamentals were actually hovering around their historical
average (Figure 8). We characterize the view prevalent in the
market at that time as one that historical average credit
fundamentals would persist into the future. Figure 10
compares the resulting rock-bottom spreads with market
spreads in May 1999. The same pattern emerges as we
exhibited in Figure 1 for October (repeated here). In
particular, a minimal or negative liquidity premium is present
in speculative grade market spreads.
Since May, spreads have backed up as the credit picture has
worsened. The question that thus arises is whether the new
levels of spreads represent good or bad value. Figure 11
compares the May-to-October rise in market spreads to the
difference in rock-bottom spreads between the historical
average and downswing credit fundamentals scenarios.
The change in investment-grade rock-bottom spreads has
been 4-7 points less than the move in market spreads.
Interestingly, 10-year swap rates, which are one measure of
the markets pricing of liquidity (see Valuing Market
Liquidity, J.P. Morgan, August 1999), are now higher by
about 10bp. Consequently, our revaluation of credit
fundamentals, plus the swap markets revaluation of the
liquidity premium, add up to approximately the backup in
market spreads.
In contrast, speculative grade market spreads have risen less
than rock-bottom spreads. As such, there is now only 4bp
of liquidity premium in BB spreads, compared to 16bp in
May, and the B-rated illiquidity premium has declined
further, from -153bp to -203bp. Evidently, while having
accounted for changes in credit fundamentals, the market
has done so in a way very different from our baseline
forecasts.
One novel development in the high yield market currently is
the stark departure between new issue and secondary market
spreads (Table 4 provides some examples). For BB-rated
issues, secondary market spreads are around 300bp at 10
years, while new-issue spreads have averaged 400bp. On
the assumption that both new and seasoned issues have the
same credit fundamentals profile, this implies an illiquidity
spread in BB-rated new issues of 104bp, which is the largest
across all rating categories. There is a similar effect in Brated issues, the difference between new-issue and
secondary market spreads standing at about 130bp.
Rock-Bottom Spreads vs. Market Spreads, October 1999

800
700
Rock-Bottom
Spread
600
500
400
300
Market Spread
200
Illiquidity
Spread
100
0
-100
-200
AAA
AA
72
6
66
90
18
72
124
34
90
Spreads
Market
Rock-Bottom
Illiquidity
BBB
BB
165
99
66
300
296
4
510
713
-203
Rock-bottom spreads calculated from the credit downswing scenario
Figure 10
Rock-Bottom Spreads vs. Market Spreads May 1999

Basis points
700
Rock-Bottom
Spread
600
500
400
Market
Spread (10 yr)
300
Illiquidity
Spread
200
100
0
-100
-200
Spreads
Market
Rock-Bottom
Illiquidity
AAA
65
6
59
AA
BBB
80
16
64
111
29
82
143
84
59
BB
275
259
16
460
613
-153
Rock-bottom spreads calculated from the historical average scenario
Figure 11
Market Spread Changes: Oct 99 versus May 99

Rock-Bottom Spread Changes: Credit Downswing vs. Historical
Basis points
100
Rock-Bottom
Spread Change
80
Market Spread
Change
60
40
20
0
AAA
AA
Spread Changes
Market
BBB
BB
Rock-Bottom
12
22
25
50
15
37
100
New York
October 25, 2001

Portfolio Research
Table 4
Indicative High Yield Spreads:

Recent New issues versus Comparable Secondary Market Issues
Secondary Issue
Spread
New Issue
Spread
Rating
LTV 09
586
Ba3/BB-
IASIS Healthcare 09
689
B3/B-
Rating
LTV 07
490
Ba3/BB-
Triad 09
509
B3/B-
Lifepoint 09
487
B3/B-
Sbarro 09
625
Ba3/BB-
Dominos Pizza 09
500
B3/B-
Unilab 09
725
B3/B-
Dynacare 06
585
B2/B+
Source: J.P. Morgan
Consequently, the B-rated illiquidity spread on new issues is

-73bp: a great improvement on the secondary markets -203bp
(Figure 10), but still not in the same risk-adjusted league as
double-Bs.
Just as we can calculate the spread implications of particular
patterns of expected future credit fundamentals, we can also
reverse the reasoning, and calculate the credit fundamentals
implied by the level of market spreads. Credit conditions were
extremely benign in early 1997 (Figure 8), as speculative
default rates fell to one-third of their long-term average, and
Moodys ratings upgrades exceeded downgrades in number
for the first time. Spreads hit an all-time low in March of 1997.
What view of the future would have justified these spreads?
Had the favorable credit conditions been expected to persist
indefinitely, the resulting rock-bottom spreads were
consistent with the typical pattern of illiquidity spreads:
about 30-50bp in high grade, small but positive for BB-rated
paper, and negligible for B-rated (see Figure 12). However, if
these favorable conditions were to last for only two years,
high-yield spreads were much lower than rock-bottom: there
was a negative payment for liquidity in the order of 40bp
implied by BB-spreads, and of about 150bp implied by Brated spreads. Evidently, spreads in 1997 priced in a
Figure 12
Illiquidity Spreads
Basis points, March 1997

100
50
0
-50
-100
-150
Early 97
conditions
persist
{forever
2 years
-200
AAA
AA
BBB
BB

page 13
substantially optimistic view of the future. Market

participants expressed concerns informally at the time. These
figures provide an explicit measure of the degree of optimism.
New York
October 25, 2001

Portfolio Research
Conclusion
Our valuation framework for credit fundamentals provides
information that is basic to any decision concerning value in
credit markets: the spread level the investor needs to
compensate for the risk of credit fundamentals.
Rock-bottom spreads are essentially a reservation price
(spread) for credit risky bonds. This kind of valuation draws
only on credit fundamentals data that are essentially external
to the market, such as default rates and recovery rates, and
on the investors risk tolerance. Thus market spreads do not
figure anywhere in the calculation of rock-bottom spreads.
The independence of rock-bottom spreads from market
spreads is what makes them useful as a valuation tool.
Comparing rock-bottom spreads with market spreads is the
way to decide whether market spreads offer good or bad
value.
Of course, if rock-bottom spreads differ from market spreads
for credit fundamentals, the implication is that the market is
using different rules to value. For example, other market
participants (in aggregate) may be employing a different
information ratio, or a different view on the future course of
defaults. We can back out such assumptions from market
spreads, as we have illustrated above.
This difference of opinion does not mean that the investors
assumptions should be brought into line with the market
assumptions. Nor does it entail that market spreads will
mean-revert to the investors valuation in the short term.
Instead, it highlights a difference of opinion or investor
circumstances, which can be expressed by taking the
different sides of a trade. As bonds near maturity, the
investor will realize a profit if his/her credit fundamentals
views were closer to reality than the markets.
We emphasize that it is the framework that is the main
message, not the baseline scenarios we have used to
illustrate the calculations and reasoning involved. Thus, our
credit downturn scenario may be thought too draconian, or
not severe enough. Either way, the framework still applies.
All that is warranted is a change in credit fundamentals
assumptions. The consequences for rock-bottom spreads
can be examined using the calculator on our Morgan
Markets website.
However, we are struck by themes that persist in the face of
large changes in credit fundamentals. For example,
significant movements in credit fundamentals produce little
movement in investment grade rock-bottom spreads,
suggesting that high-grade investors should be focusing
their efforts on the drivers of liquidity. Speculative grades
typically pay nothing, or worse, for liquidity. This may
reflect a segmented market, where greater pursuit of total

page 14
return occurs among high-yield investors. Or it may reflect

that quantitative evaluation of credit fundamentals has not
been part of the picture to date.
New York
October 25, 2001

Portfolio Research

page 15
Appendix 1:
Appendix 2:
We need to measure the correlation of movements in

issuers credit quality. To do so we adopt a variant of the
diversity score approach developed by Moodys to
analyze collateralized debt obligations. This converts a
portfolio of correlated exposures to a smaller number of
independent exposures. Moodys assume that issuers in the
same industry are highly correlated, and so diversity scores
for individual industries do not move much beyond six
equivalent independent exposures. They assume that there
is zero correlation across industries, and so add the diversity
scores of the bonds in each industry group, to arrive at a
portfolio diversity score.
Here, we look at credit-risky bonds that pay the same

coupon, irrespective of rating or maturity, and trade at
different prices to reflect their credit exposure, rather than
paying different spreads and trading at par. It is simple to
repeat our one-year-maturity calculation in this form.
Rock Bottom Spreads for long maturities
Diversity Scores
Table A1
Standard and Poors Rated Issuers, 1996

Aerospace
HiTech
Service
Leisure
Health
Building
Energy
Utility
Telecom
Transport
Finance
Insurance
Total Issuers
Diversity Score
AAA
4
3
AA
18
16
A
89
31
BBB
74
25
BB
75
28
B CCC
76
4
27
1
4
0
5
0
10
1
9
3
43
43
47
9
18
4
24
50
32
5
136
77
96
29
55
33
34
133
27
27
376
83
89
34
41
39
43
107
8
47
133
40
85
81
43
41
30
18
8
29
34
23
109
84
33
28
27
2
16
24
11
8
5
3
3
0
1
0
5
3
3
1
125
436
1013
680
495
445
29
30
53
66
63
59
54
19
We use S&Ps issuer counts by (12) industries for the entire

rated U.S. corporate market, for which we have tabulated
data (see Table A1), to arrive at our diversity score
estimates. Since Moodys partition the issuer population
into over thirty industries, their diversity score estimates
would be larger. The largest diversity score Moodys
assigned to J.P. Morgans BISTRO deals was 91 (for an Arated pool): compared to the 66 we arrive at using S&Ps 12industry classification. Adjusting our diversity scores in
this range has little impact of meaning on rock-bottom
spreads. A 50% increase in all our diversity scores, (so the
A-rated figure rises from 66 to 99), results in 10-year rockbottom spreads falling by less than four percent of their
Table 2 values.
Moodys describe their diversity score framework in their
publication Rating Cash Flow Transactions Backed by
Corporate Debt 1995 Update, by Alan Backman and Gerald
OConnor, April 7, 1995.
To fix ideas, think in terms of valuing a two-year BB-rated

bond. Whereas a one-year bond could finish the year in one
of two possible credit states (default and non-default), we
conceive eight possible credit states for a longer-maturity
bond at the end of its first year: the seven live rating
categories, and default. We know the prices of the bond in
the event that it ends up in any one of these states. Using
the probabilities of these outcomes in Figure 4, we have all
we need to calculate average credit returns and their
volatility (and thus the spread that generates an information
ratio of one-half) over the first year of the two-year bonds
life.
Credit term structures do not necessarily slope upwards: for
B- and CCC-rated credit qualities, 10-year spreads are lower
than 5-year. Here is an intuitive, if not entirely rigorous
explanation. Think of a bond as either defaulting or being
alive (irrespective of rating) at the end of one year. If it
defaults, it pays off a standard amount, irrespective of
maturity or rating. So, performance in the default scenario
does not drive the difference in spread of 10-and 9-year
bonds of the same rating category. Instead, it is what goes
on in the live scenario, which is that high grade bonds
tend to decline in credit quality over time (AAAs are an
extreme case), whereas low-grade bonds tend to improve
(from Figure 4, B-rated bonds have a 7.1% chance of
upgrade, against a 2.8% chance of downgrade). Thus, a lowgrade bond will be a better credit risk after a year if it does
not default, i.e., it will experience a capital appreciation in the
live scenario (over and above pull-to-par), and so it can
command a lower spread. Similarly, a high-grade bond will
be a worse credit on average in the live scenario, and so it
requires an increasing spread as maturity increases.
Throughout these calculations, we assume that the
government curve is flat and does not change over time,
which is, of course, wildly unrealistic but greatly simplifies
and shortens calculations. However, we are looking at
credits outperformance of governments, which is analogous
to hedging away government exposure. As a result, the
actual dynamics of the government market do not greatly
affect rock-bottom spreads. Pricing credit risk is the context
of a mean-reverting government curve, exhibiting the
volatility of the last 10 years, lowers 10-year BB rock-bottom
spreads by 3bp.
New York
October 25, 2001

Portfolio Research
Appendix 3:
Translating default, drift and activity forecasts

Rating categories have differed in the volatility of their actual
transitions, and their sensitivity to the overall incidence of
rating changes. We account for this by inferring a forecast of
each rating categories downgrades from a historical
regression on aggregate downgrades, and we repeat this
process for upgrades and speculative defaults. The slopes of
these regressions (shown in Table A3) increase as credit
Table A3
Transition Matrix Adjustments

Downgrades
Slope Change
AAA
AA
A
BBB
BB
B
CCC
0.38
0.79
1.00
0.89
1.52
1.13
0.7
1.4
1.7
1.6
2.7
2.0
Upgrades
Slope Change
Speculative Defaults
Slope
Change
0.25
-0.01
1.51
1.96
2.18
2.89
0.00
0.00
0.01
0.07
0.36
1.31
3.71
-0.4
0.0
-2.6
-3.4
-3.8
-5.1
0.00
0.00
0.01
0.12
0.69
2.50
7.05
quality declines, testifying to the different sensitivities of

rating categories to the overall environment. The table also
shows the changes in downgrade rates that result from our
credit downturn scenario, which pushes aggregate
downgrades from 14.25% to 16% yearly (Table 3). For
example, the implied forecast of AA downgrades is an
increase of 1.4 percentage points. This results from the
forecast aggregate downgrade figure exceeding the 14.25%
historical average figure by 1.75%, multiplied AA-bonds
sensitivity of 0.79 (1.75*0.79=1.4).
Our last step is to propagate these changes among
downgrades by one, two, or more notches. We distribute the
increase in downgrades according to the shares of different
notches in Moodys historical transition matrix (see Figure 4).
For example, downgrades of AAs to A have historically
accounted for 94% of all downgrades (9.6/(9.6+0.3+0.15+0.15),
from the second row of the transition matrix). Accordingly, the
AA-to-A downgrade rate in our forecast transition matrix will
be 9.92% (9.6%+0.94*1.4%).

page 16
New York
October 25, 2001

Portfolio Research

page 17
August 1, 2001
Portfolio Research
Rock-bottom spread mechanics

A bonds rock-bottom spread prices its credit exposure
We explain with examples what goes into
calculating rock-bottom spreads, emphasizing:
o
o
how credit risk is accounted for

the key role of views on default and
recovery rates, and how to formulate them
the minor effects of interest rates and

recovery rate uncertainty
A new rock-bottom spread calculator on

Morganmarkets lets you replicate and further
dissect the examples
how the combination of your investment performance goals

and a bonds credit fundamentals determine its rock-bottom
spread. Then we explain how the rock-bottom spread is
affected by each element of credit fundamentals: coupons,
maturity, recovery rates, default rates, and so on. We also
describe a simple framework for translating your own views
of future credit fundamentals trends into rock-bottom spread
numbers that you can compare with market spreads.
Figure 1
Rock-bottom spreads versus market spreads (bp)
Rock-bottom
exceeds
market spread
900
800
700
600
Corporate bonds have two principal drawbacks relative to

swaps or government bonds: greater credit exposure, and
lower liquidity. Investors need to assess whether market
spreads pay enough for these drawbacks. This task is more
manageable if the exposures can be translated into spread
terms. JP Morgans rock-bottom spread framework does
just this for credit exposure.
Rock-bottom
less than
market spread
500
400
300
200
Positive
liquidity
spread
100
Rock-bottom spreads are built around the idea of a

reservation price: the highest price at which you can buy an
asset and remain consistent with your investment
performance goals. A bonds rock-bottom spread is the
amount you need to be paid for bearing its potential for
downgrade or default, and expected recovery rate, taking
into account your ability to diversify these risks. If the
bonds market spread is below its rock-bottom spread, on
average it will not deliver sufficient return for the credit risk
it entails. Similarly, the difference between its market
spread and rock-bottom measures how much you are getting
compensated, if at all, for its lower liquidity. Figure 1
illustrates.
In several other research pieces, we have used the rockbottom spread framework as a common yardstick for
disparate asset classes such as high yield and investment
grade corporate bonds and emerging markets sovereigns.
Similarly, the approach makes it simple to place a value on
credit-driven features of individual securities, such as call
provisions in high yield bonds, and coupon step-up clauses
triggered by rating changes.
The mechanical steps involved in calculating rock-bottom
spreads are explicitly laid out in this note. We first show
AA
BBB
BB
Market
Rock-bottom
spread
spread
Spreads over US Treasuries curve for (duration-weighted) averages of 510yr senior, unsecured bonds in the JP Morgan High-Yield index (BB and
B-rated), and liquid investment grade bonds. Market spreads are as of
June 1, 2001. Rock-bottom spreads incorporate the negative credit view
for the next 12 months that is described in the last section of the paper.
Recovery rates on defaulted bonds are assumed to be $35 per $100 of
principal, in line with current traded prices of defaulted debt.
Of course, this is scarcely vacation reading. Rather, it is

intended as a reference that will make it possible to take
apart the rock-bottom spread calculations in our
publications, and that can now be carried out painlessly
using our new web calculators. All but a few of the
calculations in this research note used these calculators. In
the text, we provide the recipe that will enable you to
replicate each calculation. An accompanying spreadsheet,
available from our website [10] 1 , demonstrates how to build
a simple rock-bottom spread calculator. Thus armed, you
will be able to take apart and rationalise any rock-bottom
spread valuation.
1
Numbers is square brackets refer to the publications listed on p.12
New York
October 25, 2001

Portfolio Research

page 18
Valuation
Table 1
Valuing an asset means determining a reservation price at

which you are prepared to buy it. The reservation price is
the highest price consistent with achieving your investment
objectives, given your views of how the asset will perform.
Thus, implicit in any valuation of a financial asset are:
Moody's 1-year default rates
the prices at which you believe the asset can be sold in

each possible future scenario;
your view of the relative likelihood of each scenario;
a performance target, typically relating to the assets
risk and return
Averages, 1983-2000
AAA
AA
A
BBB
BB
B
CCC
0.00%
0.03%
0.00%
0.18%
1.52%
7.46%
29.21%
Performance criterion
We will now explain the role of each of these components in
the rock-bottom spread valuation framework.
Credit scenarios
For bonds that pay cash, the future scenarios are quite simple
to lay out, because we have an anchor at the maturity date.
For example a one-year (annual-paying) 8% bullet promises
$108. Of course, the key word here is promises, because
the bond can also default. The payout in the event of default
is uncertain, and will be the result of a complicated
liquidation or restructuring process. For concreteness, we
shall assume that the bond is worth an average of $45 per
$100 of principal, in line with the historical average prices of
bonds that have just defaulted2 . So, at maturity, two credit
scenarios are possible (see Figure 2).
Figure 2
Credit scenarios for a 1-year 8% bond
No default
$108
Default
$45
Today
In 1 year
The value you place on the bond will depend on how likely
you view each scenario to be. For example, your view
could be in line with the frequency of default that has been
observed historically (see Table 1). We shall use these
figures to illustrate how to calculate rock-bottom spreads,
but it is worth stressing that there is nothing sacred about
them. They simply describe past experience, and have no
automatic claim to represent your view of future credit
conditions. They are just one possible set of input values
that really only should be used if you have no strong
feelings about credit trends.
2
Of course, this is a simplification; recovery rates are anything but certain once a
bond has gone into default. The full rock-bottom spread calculation prices in
uncertainty about recovery rates.
The performance criterion behind rock-bottom spreads is

based on the information ratio. In general, an assets
information ratio measures its return relative to a
benchmark, adjusted for risk. It is just the difference
between the expected returns of the asset and benchmark,
divided by the assets return volatility around the
benchmark, or tracking error. The higher the information
ratio, the better the asset is expected to perform.
Requiring your investments to attain a target level of the
information ratio is a sensible performance criterion. It
corresponds to setting a target rate of return on capital.
Taking risk around your benchmark costs capital in one
form or another, and a low information ratio implies that
you are getting a low return on capital. Here, we set the
target annual information ratio at 0.5, which has become
something of a standard in the investment management
industry. In particular, we can point to investment
strategies, such as active management of global government
bond funds, which have achieved this target in the past.
Why pursue a strategy that produces a lower information
ratio than the available alternatives?
From target information ratio to rock bottom

How does setting a target level for the information ratio
enable you to put a value on credit exposure? Via its
dependence on returns, an assets information ratio also
depends on its current price: the higher the price, the lower
the information ratio. So there is some price at which the
information ratio will just equal 0.5. This price is precisely
the value you place on the asset. It is the highest you can
afford to pay and still satisfy your performance criterion. In
the context of valuing credit exposure, this defines the rockbottom price, from which the rock-bottom spread follows
via a conventional price-to-yield calculation.
To calculate rock-bottom spreads, we thus need to assemble
the components of the information ratio provided by credit
exposure. In our case, the relevant benchmark is an
investment on which there is no chance of default, which we
shall call a government bond. Our 8% bonds credit return
New York
October 25, 2001

Portfolio Research
in a scenario (default or no default) is the excess of its return

over the government bond:
pricein
current
credit return
= scenario
price
in scenario
current price
govt

return
So, the information ratio we are interested in is:

average credit return
information
=
credit return volatility
ratio
The numerator of the information ratio averages across the

credit return scenarios, while the volatility in the
denominator is the standard deviation of credit returns
across the scenarios.
The information ratio depends on the price you pay for the
bond today, because credit returns depend on the price
today. Since we know the prices in each future scenario
(see Figure 2), and the probabilities of each scenario, the
only unknown quantities in this equation are the information
ratio and the current price. Once we set the information
ratio at its target level, there is only one unknown, the
current price. The value of the current price that satisfies the
equation is the rock-bottom price: it is the price consistent
with the target information ratio, under the assumed credit
conditions
As a practical matter the rock-bottom price could be solved
from the information ratio equation by trial and error.
However, it is easier and more revealing to reexpress this
equation as a definition of the rock-bottom price.
First, we deal with the average credit return across
scenarios. To calculate this, you multiply the credit return
in each scenario by its probability, and sum across
scenarios. We assume for the moment that the government
return does not change across scenarios. So, the price in
scenario term is the only element of the credit return that
changes from one scenario to another. The average of the
government return across scenarios is just the government
return, and the same goes for the current price.
Consequently, the average credit return is:
average
credit =
return
average price
across scenarios
current price
govt
1 +
return
A similar rule holds for the volatility of credit returns, which

is just
credit
return
volatility
price volatility
across scenarios
current price

page 19
Why is the (1+govt return) term absent from volatility?

Since it has the same effect in each scenario, its level does
not affect the range of variation across scenarios, which
volatility measures. Why is the current price present in the
denominator? It would not be there if we were just looking
at the volatility of profit and loss (future price minus current
price). But returns are scaled P&Ls: the more you pay
currently, the smaller the proportionate return (in absolute
terms) resulting from any price movement, so, the smaller is
their range of variation, and volatility needs to be scaled
accordingly.
Now we can put together these two pieces of the
information ratio, and a little bit of algebra expresses the
current price in terms of average future prices and their
volatility, the information ratio, and the government return:
current
=
price
average price
across scenarios
(1
information price volatility
ratio across scenarios

+ govt return)
The rock-bottom price is simply the price that delivers the

investors target information ratio. If we now set the
information ratio at the target level of 0.5, we have a rule for
the price that will deliver that target, given expected credit
returns and their volatility, and given government returns:
Rock
Bottom =
price
average price
across scenarios
(1
+ govt return)
target
price volatility
n
informatio
* across scenarios
ratio
(1
+ govt return)
Notice that the current market price of the bond is absent

from this definition. This is exactly as it should be if rockbottom prices are to offer an independent measure of value
against which market prices can be compared.
The final step is to translate the rock-bottom price into a
rock-bottom spread. First, we calculate the yield to maturity
on the bond that is implied by its rock-bottom price and
cashflows. Then we calculate the yield of the identical set
of cashflows, using the government curve. The difference
between these two yields is the rock-bottom spread over
government rates.
Calculating Rock-Bottom Spreads

We now proceed to use the rock-bottom price equation to
value generic bullet bonds. We distinguish bonds by their
coupon, maturity, seniority, and senior credit rating of their
issuers. To keep the volume of numbers manageable, we
stick to the broad or 8-state rating categories listed in
Table 1, instead of the more detailed 18-state
classification (BBB+, Baa1, etc) now used by the rating
agencies. Most of the calculations that follow can be
New York
October 25, 2001

Portfolio Research
replicated on the 8-state version of our web calculator3 , by

entering the appropriate input values. Each calculation is a
small variation on a basic set of inputs:
Baseline inputs for rock-bottom spread calculations
Default/Downgrade View
Recovery rate average
Historical
0.45
Recovery rate volatility

Information ratio
0
0.5
Diversity score
70
Government curve
Flat 6%
Coupon
8%
The terms Diversity score and Default/Downgrade View will be
clarified below.

page 20
issuers. As a practical matter, it is easiest to calculate the

impact of portfolio diversity when issuers fortunes are
independent of each other (and horrendously difficult if they
are correlated). Consequently, we use a variant of Moodys
diversity score measure to translate a portfolio of
correlated issuers paper into a (smaller) number of
independent issuers that would provide the same degree of
diversification.[2,3] The price volatility of a diversified
portfolio is simply the price volatility of a single bond,
divided by the square root of the diversity score. As a
result, the rock-bottom price for a portfolio is
Rock
Bottom
price
price volatility
Target
across scenarios
average price
information *
across scenarios
ratio diversity score
(1
+ govt return )
Rock bottom: a single 1-year bond

Table 2 assembles the elements of the rock-bottom
calculation for the 8% one-year bond, on the assumption
that it is rated BB. The average future price is $107.04, and
the volatility of future prices is $7.71. Assuming a
government return or annual yield of 6%, this translates into
a rock-bottom price of $97.35, equivalent to a spread of
494bp. By historical standards, this is quite high for a BBrated bond. It is correct for an investor whose entire
portfolio is invested in a single issuer: for this type of
investor, a spread of less than 494bp will result in an
information ratio of less than 0.5. However, investors
typically hold more diversified portfolios, the result of
which is to lower the volatility of the credit returns they
face. As a consequence, a lower spread will suffice for their
target information ratio.
Figure 3 traces the effect of diversification on the rockbottom spread of a BB-rated portfolio. As diversification
increases, rock-bottom spreads initially drop precipitously,
but at a diversity score of 20 or so, the curve has all but
flattened out. This is fortunate, since calculating diversity
scores is far from a precise science. The Figure shows that
it does not matter materially whether the diversity score is
50 or 100: the resulting difference in BB rock-bottom
spreads is 17bps.
Figure 3
Rock Bottom Spreads for BB-rated bonds held in a

diversified portfolio
500
450
Table 2
400
Average Price and Price Volatility for a 1-yr BB-rated

Bond
350
Deviation
from
Average
*
Probability
Price
Price
Scenario
Price in
Scenario Probability
Squared
Deviations
*
Probability
300
250
200
150
No Default
Default
108
45
98.48%
1.52%
106.36
0.68
Average
107.04
0.96
-62.04
0.90
58.51
100
0
107.04
0.5 * 7.71
20
30
40
50
60
70
80
90
100
Diversity score
Volatility
Rock
Bottom =
price
10
7.71
97.35
0.06
To produce these figures with the 8-state rock-bottom spread calculator,

use the Baseline Inputs, but set the diversity score equal to 1.
Incorporating diversification
The entire US High Yield corporate sector offers a diversity

score of about 70. This means that, although there are
approximately 1000 rated speculative-grade issuers they
only provide the same diversity as would 70 issuers with
independent asset values. 4 A fully diversified investor
would thus face a rock-bottom spread of 141bps for a BB
bond.5 Less-diversified investors would require greater
rock-bottom spreads for one-year BB bonds.
Portfolio diversification results from (the lack of)

correlation of the underlying asset values of the bond
4
3
The calculator is accessible on the Credit Page of MorganMarkets.com, and is

described in reference [1]
The Appendix to [3] contains a fuller discussion of diversity scores

The rock-bottom price in Table 2 becomes (107.04-0.5*7.71*70)/1.06 =
100.55
5
New York
October 25, 2001

Portfolio Research
Interpreting rock-bottom spreads

The rock-bottom price has two components. The first is the
breakeven price:
Break
even =
price
average price
across scenarios
(1
+ govt return)
so called because if this is what you pay for the bond, then
you will earn the same return on average as you would by
investing in government bonds. Consequently, your
average credit return will be zero.
Only investors who are indifferent to risk would settle for
breaking even. If you are risk-averse, you will refuse to pay
as much as the breakeven price, because of the uncertainty
credit exposure presents. Via its second component, the
rock-bottom price requires a discount from the breakeven
price for this risk:
Rock
Bottom =
price
Break
even
price
Risk
Discount
In its current form, the risk discount is driven by price

volatility associated with the uncertainty of the issuers
rating one year from now, and diversification. Later, we
shall also include the effects of uncertain recovery rates,
without changing the basic form of the rock-bottom price.

page 21
anomaly, which we could have removed by smoothing out the default

probability estimates, rather than using the raw historical numbers, as we
have done for simplicity.
Table 3 shows how one-year rock-bottom spreads

decompose into a breakeven spread (corresponding to the
breakeven price) and a risk premium (corresponding to the
risk discount component of the rock-bottom price). Each
rating category entails different default probabilities, and
therefore different average future prices and breakeven
spreads. The risk premia rise as the probability of a default
rises, because this causes price volatility to rise.
Longer-maturity bonds
To this point, we have applied the principle that a bonds
value today is driven by its value in the future to derive
rock-bottom prices and spreads for one-year bonds. We can
now use these one-year bond values to value two-year
bonds. One year from now, a two-year 8% BB bond can
finish up in default, as before. Alternatively, it can finish
the year as a one-year 8% bond rated AAA, AA, A, BBB,
BB, B, CCC. Its value with one year to go will be different
according to its rating, as we have calculated in Table 3,
because its chance of going into default in its last year will
differ in each case. So, we need to expand the no-default
category to these possibilities, as Figure 4 shows.
Figure 4
Price scenarios for a two-year bond

2 yrs to go
While we have thus incorporated risk aversion in the rockbottom price, we have not done so by stating outright how
risk-averse investors are. We have simply reasoned that
they should demand from credit a risk discount that brings
its performance in line with what they apparently demand
from other investment strategies. This led us to a target
information ratio of 0.5.
BB
1 yr to go
AAA 109.89
AA 109.80
109.89
A
BBB 109.63
BB 108.55
104.52
B
CCC 90.91
45.00
D
Maturity
AAA 108
108
AA
108
A
BBB 108
108
BB
108
B
CCC 108
45
D
Table 3
Valuing 1-yr 8% Bonds
AAA
AA
A
BBB
BB
B
CCC
Breakeven
Spread
0
2
0
11
95
483
2177
Risk
Premium
0
7
0
16
46
107
249
Rock
Bottom
Spread
0
9
0
27
141
590
2426
Rock
Bottom
Price
101.89
101.80
101.89
101.63
100.55
96.52
82.91
To produce the rock-bottom spreads with the 8-state rock-bottom spread

calculator, use the Baseline Inputs. Breakeven spreads result by setting the
information ratio equal to zero.
AA spreads are higher than A spreads because the historical AA-rated oneyear default rate is higher (see Table 1). This derives from the fact that
one issuer rated A3 at the start of a calendar year defaulted within that year
(DFC, in 1989). It does not mean that Aas are more risky than As, as
default frequencies over longer horizons show [4]. It is a statistical
It remains to attach probabilities to these events , namely that

one-year from now, our two-year BB bond will be rated
AAA, or A, etc. Again, we use historical data for
illustration. Table 4 shows the frequency of changes in
ratings by Moody over the last two decades. If we treat
these as the probabilities of future changes in credit quality,
then we have all the ingredients to calculate rock-bottom
prices and spreads for two-year bonds. The relevant
probabilities for BB bonds are in the fifth row. Table 5
details the calculation of the rock-bottom price of 100.67,
which translates into a rock-bottom spread of 162bps.
As noted above, this calculation penalises bonds for their
volatility. However, it is only credit volatility that is
considered, by which we mean the variation in the future
value of the bond as its future credit quality varies (see
Figure 4). Market price volatility does not enter into the
New York
October 25, 2001

Portfolio Research
picture, which is in keeping with our aim of pricing the

credit component of the bond.

page 22
Spread term structures
Table 4
Moody's 1-yr credit migration rates

AAA
AA
AAA
89.20%
1.03%
AA
9.69%
89.31%
A
1.08%
9.14%
BBB
0.00%
0.37%
BB
0.03%
0.09%
B
0.00%
0.02%
CCC
0.00%
0.00%
D
0.00%
0.03%
A
BBB
BB
B
0.04%
0.04%
0.03%
0.01%
2.48%
0.29%
0.03%
0.06%
90.97%
6.24%
0.60%
0.25%
5.57%
86.96%
5.59%
0.58%
0.72%
5.15%
82.94%
6.43%
0.21%
1.09%
8.67%
82.06%
0.01%
0.05%
0.62%
3.14%
0.00%
0.18%
1.52%
7.46%
CCC
0.00%
0.00%
0.00%
1.12%
2.87%
6.77%
60.03%
29.21%
It may seem counterintuitive that B- and CCC-rated rockbottom spreads fall with maturity, since a longer maturity
means a greater chance of losing the principal. Actually, the
falling low-grade credit curve is no more surprising than the
rising high-grade credit curve, and both emanate from the
same source.
Figure 5
Cumulative Default Probabilities

8%
BBB
7%
Table 5
Average Price and Price Volatility for a 2-yr BB-rated
Bond
Scenario
Price in
Scenario
Probability
Price
*
Probability
109.89
109.80
109.89
109.63
108.55
104.52
90.91
45.00
0.03%
0.03%
0.60%
5.59%
82.94%
8.67%
0.62%
1.52%
0.04
0.03
0.65
6.13
90.03
9.06
0.57
0.68
Deviation
from
Average
Price
2.69
2.61
2.69
2.43
1.36
-2.68
-16.28
-62.19
Squared
Deviations
*
Probability
6%
5%
4%
3%
A
2%
1%
AAA
AA
A
BBB
BB
B
CCC
D
0.00
0.00
0.04
0.33
1.52
0.62
1.65
58.76
AA
AAA
0%
0
10
90%
CCC
80%
70%
60%
50%
Average
Rock
Bottom
price
40%
107.19
7 .93
70
7.93
0.95
Volatility
Diversified
30%
BB
20%
107.19 0.5 *
=
1 + 0.06
10%
= 100.68
Time horizon (years)

0%
0

use the Baseline Inputs
The recipe for 3-year bonds uses the prices calculated for 2year bonds as input, and so on. In this way, we can trace out
an entire credit- and term-structure of rock-bottom spreads,
as shown in Table 6.
Table 6
Rock-bottom spreads by rating and maturity

Based on 8% annual coupon bond
Maturity
AAA
AA
A
BBB
BB
B
CCC
10
0
9
0
27
141
590
2426
1
9
3
33
162
590
2088
1
9
6
39
179
584
1820
2
9
11
50
204
563
1452
3
10
17
60
219
539
1228
5
13
24
72
230
506
1037

use the Baseline Inputs
10
Figure 5 shows the cumulative default probabilities,

derived from the figures in Table 4.which indicate the
chance that a bond will have defaulted by a given number of
years from now. While these probabilities must rise as the
time horizon lengthens, they do so at an increasing rate in
the case of bonds now rated investment grade, but at a
decreasing rate for B and CCC-rated bonds. For example,
the probability that a CCC-rated issuer will default within 5
years is 71%, while the 10-year probability is only 10%
higher, at 81%. This dramatic difference occurs because
those (currently) CCC-rated bonds that survive 5 years will
probably be rated much higher, and so will enjoy a much
lower default probability from then on.
Now, say you require 1452bp of spread to bear 5-year CCC
exposure, as in Table 6. Would you be prepared to accept
this spread to extend your exposure for a further 5 years,
that is, to 10 years? Over these back 5 years, you only have
10% extra chance of default, or about 2% a year, well below
B-rated risk, for which you require less than 600bp, again
according to Table 6. So, in fact, the 1452bp is more than
you would settle for. You would accept a lower spread
because the risk you are taking in an average year has been
diluted. Granted, the chance of a CCC bond reaching 5
New York
October 25, 2001

Portfolio Research
years is small, but that is reflected in the 1452bp, and should

you get to that point, you will then be holding a more
valuable bond.
Exactly the reverse reasoning explains why higher grade
credit curves slope upwards. Because cumulative default
probabilities rise at an increasing rate, the back 5 years of
the life of a 10-year A-rated bond will be more risky than
the initial 5 years, and warrant more spread accordingly.
A wrinkle: the government benchmark

So far, we have been comparing bonds with credit risk to an
unspecified government benchmark, which we have
assumed to pay an annual return of 6%. Talking about the
government return is somewhat vague. What we are really
after is a comparison of the return of the credit bond and a
government bond promising the same cashflows. Our 6%
constant return assumption is tantamount to the belief, held
with certainty, that the government curve will be flat until
the maturity of the longest bond we have priced.

page 23
government rates moves B-spreads by about 9bp. Similarly,

whether we use a flat 6% curve, or the current government
forward curve, or assume rates vary in line with a standard
term structure model, single-B spreads are affected to the
tune of 13bp or less.
Government rates are not significant because, to a first
approximation, they will affect the prices of government and
corporate promises of cashflows similarly. Just as (forward)
government rates are used to discount the known cashflows
of government bonds, they are used to discount the expected
cashflows of bonds with credit risk, as the rock bottom price
definition above shows.
Simulating random government rates does not appear worth
the trouble, unless we are valuing bonds where
contingencies on government interest rates are important,
for example, high-grade callable bonds. Accordingly, our
web calculators proxy future one-year rates using the
current government forward curve.
Table 7
Rock bottom spreads under alternative government

yield curve assumptions (bp)
Rockbottom
spread
AAA
AA
A
BBB
BB
B
CCC
6% Flat
Curve
5
13
25
72
230
506
1036
Difference from
6% Flat Curve
0% Flat
Curve
0
+1
+1
+5
+19
+54
+94
Forward
Curve
0
0
-1
0
+1
+3
+14
Random
Curves
0
0
-1
+1
+4
+13
+46

use the Baseline Inputs, with the following amendments: substitute 0% for
the government return to produce the spreads behind the second column.
Select the Research paper forward curve option on the 8-state calculator
for the third column. Here, we are assuming that each of the one year
forward rates (of Feb 26, 2001) is the actual rate in the relevant year. These
forward rates ranged from 4.1% at one year, to 6.7% at 30years.
The "Random Curves" calculations cannot be reproduced with the 8-sate
calculator. They simulate interest rate fluctuations around this forward
curve using the Cox-Ingersoll-Ross model, assuming the volatility of the 1year rate to be 30%, in line with historical estimates.
While this is a wildly unrealistic view of government rates,

its effects on rock-bottom spreads are not terribly serious.
Table 7 shows that rock bottom spreads are much the same,
irrespective of the way we account for government rates.
For example, single-B 10-year rock-bottom spreads differ
by only 54bp, whether the government curve is assumed flat
at 6% or flat at 0%. In other words, a 100bp inaccuracy in
Coupon effects
While changes in government rates seem to be of secondary
importance for rock-bottom spreads, the level of the bonds
coupon can have a significant effect. Table 8 shows that
rock-bottom spreads fall at an increasing rate as the bonds
promised coupon falls. For example, an 8% BB bonds
rock-bottom spread is 37bp lower than that of a 16% bond.
However, if the coupon is lowered another eight percentage
points, to zero, the rock-bottom spread falls by a further
73bp, to 157bp. The source of this effect is very different
from the shape of the yield curve, which drives coupon
effects on government bonds. A bond that pays a low
coupon will have to command a low price, so that it can pay
an expected return in excess of governments. If the bond
defaults, we assume that the investor receives the same
recovery rate, irrespective of coupon.6
Table 8
10-year rock bottom spreads according to promised
coupon
0%
AA
Promised Coupon
8%
16%
10
13
14
BBB
BB
58
157
72
230
80
267
266
506
635

use the Baseline Inputs but set the coupon to the specified level.
6
This corresponds to the way recovery rates are measured, i.e. as an amount per $100
face amount, irrespective of missed coupons (see [5]). It is also entirely in line with
the markets approach. Some 350 defaults over the last 20 years show no relationship
between the prices of defaulted debt, and the size of the cashflows on which the
default occurred.
New York
October 25, 2001

Portfolio Research
Consequently, the capital loss on a low-coupon bond will be

less than on a high-coupon bond. This milder loss in
default mean that compensation in the no-default scenarios
can be correspondingly lower, and this is reflected in the
lower spreads for low-coupon bonds.
This result does not necessarily mean that low-coupon
bonds are a better deal than high-coupon bonds. Market
spreads may over- or undercompensate for this coupon
effect. As in other cases, the appropriate route to
determining which offers better value is to calculate their
rock-bottom spreads, which correctly price the coupon
effect, and then determine which pays the greater surplus
over rock-bottom

page 24
Average recovery rates

Although the uncertainty of recovery rates has little impact
on recovery rates, the same cannot be said for the average
level of the recovery rate. Table 10 shows the effect of
average recovery rates on 10-year rock-bottom spreads, in
relation to the $45 assumption we have been using. As one
would expect, the effect is larger the lower the credit rating.
A rough rule-of-thumb is that each $1 increase in the
average recovery rate raises rock-bottom spreads for BBBs,
BBs, and Bs by 1,4, and 10bps, respectively.
Table 10
Rock-bottom spreads under alternative recovery rate

assumptions
Basis
points
per $

(per $100 principal)
A wrinkle: uncertain recovery rates

To this point, a default has been assumed to result in an
immediate payout of exactly $45. In practice, there has
been a wide range of variation in recovery rates. One study
estimates the volatility of recovery rates to be about $23 per
$100 of principal. Obviously, the greater the uncertainty
about recoveries in the event of default, the higher should be
the spread investors need to be induced to take credit risk.
Not surprisingly, uncertainty of recovery rates affects rockbottom spreads through the volatility of credit returns.
Perhaps more surprising is the fact that uncertainty of
recovery rates has a small impact on the values of credit
instruments, shown in Table 9. Although large in dollar
terms, the effect of recovery rate uncertainty on rock-bottom
prices is small because it is scaled by the probability of
default and the diversity score 7 .
Table 9
Increase in rock-bottom spreads from maximum

possible recovery rate uncertainty
Maturity
10
AAA
0
0
0
1
AA
2
2
2
2
A
0
1
1
2
BBB
4
5
5
5
BB
13
14
14
15
B
32
33
33
33
CCC
96
96
88
69
use the Baseline Inputs, but set the recovery volatility equal to 0.5.
7
To this point, each scenario has resulted in a single outcome, as depicted in Figure
2. Now, the default scenario comprises a range of outcomes, described by the
recovery volatility. The impact of this is to raise the contribution of the default
scenario to credit return volatility. Specifically, the square of credit return volatility is
raised by the default probability multiplied by the square of recovery volatility. In
terms of the one-year BB example of Table 2, the square of credit return volatility
becomes (7.71)^2 + (0.0152)*(23^2)=59.44+8.05=67.49. So credit return volatility
rises to $8.22 from $7.71, or $0.51, as a result of accounting for recovery volatility.
The effect on the rock-bottom price is considerably less, as this figure is scaled down
by the diversity score, the information ratio, and the government discount factor.
AAA
AA
A
BBB
BB
B
CCC
$0
8
21
41
122
407
990
2848
$20
6
18
34
100
325
751
1761
$45
5
13
25
72
230
506
1036
$70
3
8
15
46
143
304
579
0.1
0.2
0.4
1
4
10
32

use the Baseline Inputs, substituting each columns recovery rate average
(expressed as a decimal).
Forward-Looking Rock-Bottom Spreads

The relevant credit fundamentals for valuing a security are
those expected in the future, which may or may not conform
to the historical averages used in the illustrative calculations
above. For example, in an economic downswing, default
rates and the general direction of changes in credit quality
can depart markedly from the average. Similarly, recovery
rate expectations may differ from historical averages, as the
industrial composition of the market changes. Meaningful
rock-bottom spreads require taking a view on these elements
of credit fundamentals for each year in the remaining life of
the securities being valued.
Of all the elements of a view on credit fundamentals, the
most burdensome is the frequency of expected changes in
ratings and defaults. These are summarized in a credit
migration matrix such as Table 4. In this eight-state form,
there are 49 independent numbers. In principle, a forecast
of each of these numbers is needed for each year until
maturity, and it would be an enormous task to make these
forecasts from scratch. However, for most purposes, a
simple approximation seems to suffice.
The historical record suggests that movements in ratings are
highly correlated across rating categories. For example,
Figure 5a charts the frequency of downgrades by Moodys
among investment grade issuers. The experiences of the
New York
October 25, 2001

Portfolio Research

page 25
Figure 5
Aggregate credit migration

(a) Downgrade rates (%)
(b) Default rates (%)

100
20
25
Downgrade
rate
Ba, B
Default
rate
18
Caa
Default
rate
90
Caa
16
80
14
70
12
60
20
Baa
All Investment Grade
Aa
15
50
10
All Speculative Grade
10
40
30
20
5
A
Ba
10
0
0
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
1980
2000
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
Figure 6
Change in annual credit migration probabilities associated with a 1% increase in

(a) Aggregate investment grade downgrades
%
AAA
AA
A
BBB
BB
B
CCC
1.5
AAA
-0.756
0.184
0.003
-0.001
-0.002
-0.001
0.000
AA
0.687
-1.271
0.123
-0.004
-0.005
-0.003
0.000
A
0.067
1.023
-1.069
-0.089
-0.040
-0.015
-0.024
BBB
0.000
0.032
0.820
-0.843
-0.382
-0.051
-0.037
BB
0.002
0.016
0.097
0.740
-0.415
-0.481
-0.105
B
CCC
0.000 0.000
0.016 0.000
0.025 0.001
0.121 0.010
0.507 0.031
-0.375 -0.282
-0.206 -3.354
(b) Aggregate speculative grade defaults

D
0.000
0.000
0.000
0.065
0.306
1.207
3.725
%
AAA
AA
A
BBB
BB
B
CCC
4.0
AAA
0.604
0.012
-0.003
-0.003
-0.001
-0.001
0.000
AA
-0.549
-0.580
-0.158
-0.021
-0.004
-0.002
0.000
BBB
0.000
0.017
0.328
0.142
-0.275
-0.036
-0.098
BB
B
-0.002 0.000
0.008 0.008
0.039 0.010
0.282 0.046
-0.942 0.730
-0.343 -1.266
-0.276 -0.544
CCC
0.000
0.000
0.001
0.004
0.045
0.125
-2.543
D
0.000
0.000
-0.002
0.037
0.476
1.533
3.524
3.0
AAA
A
-0.053
0.535
-0.215
-0.487
-0.029
-0.010
-0.062
BB
1.0
AA
B
2.0
CCC
0.5
BBB
1.0
0.0
0.0
-0.5
-1.0
-1.0
-2.0
-3.0
-1.5
Up 3
Up 2
Up 1
No
change
Down 1
Down 2
Down 3
Down 4
Down 5
AA, A, and BBB sectors have been very similar. They track
the aggregate rate of downgrades closely, with correlations
of 71%, 94%, and 89% respectively. Indeed, most of the
systematic change in the pattern over the cycle of
investment grade credit migration seems to be a shift
between the unchanged rating state, and downgrades. The
pattern for upgrades is much more erratic.
Similarly, there has been a strong correlation between the
aggregate speculative grade default rate, and the default and
downgrade experience of individual speculative grade rating
categories. Figure 5b exhibits the pattern for BB, B, and
CCC default rates, whose correlations with the aggregate
default rate are 78%, 81% and 33%, respectively. The
correlation of the BB downgrade rate with the aggregate
default rate has been 37%, while the corresponding figure
for Bs is 22%.
Up 5
Up 4
Up 3
Up 2
Up 1
No
change
Down 1
Down 2
Down 3
These historical relationships suggest that not much will be

lost by taking a view, for each year you need to forecast, on
at most two aggregates:
investment grade downgrades

speculative grade defaults
We then need to translate these aggregate forecasts into

forecasts for each element of the credit migration matrix.
For example, say you forecast next years aggregate
investment grade downgrade rate to be 11.5%, which is 4%
higher than the 7.5% historical average. How does this
translate into a forecast for, say, downgrades from A to
BBB?
The process is similar to using an individual stocks beta to
forecast its return, based on a forecast of the market. The
slope of a regression of the history of A-to-BBB
New York
October 25, 2001

Portfolio Research

page 26
downgrades on the history of aggregate investment grade

downgrades is 0.82. This slope tells us how much A-toBBB downgrades will deviate from their historical average
when aggregate downgrades exceed their historical average
by 1%.
grade rows result from adding four times the bottom three
rows of Figure 7b to Table 4. (There is nothing to stop the
resulting probabilities from being negative. In this case,
probability is reallocated from the nearest positive
probability, to remove the negative numbers.)
From Table 4, the historical average for A-to-BBB

downgrades is 5.57%. Since you have forecast aggregate
investment grade downgrades to be 4% above their
historical average, this implies you expect the A-to-BBB
migration probability to be 0.82 x 4%=3.28% above its
historical average, or 8.85%.
We would expect rock-bottom spreads to be higher as a

result of this view, and to be more elevated, the longer the
credit downswing is expected to last. These views are
confirmed by Table 11, which shows the consequences of
expected downswings lasting through 2001 03. In each
case, the credit migration probabilities are assumed to revert
to historical levels (i.e., those in Table 4) after the last
downswing year.
We have estimated these regression slopes for each element

of the credit migration matrix, using each of the two
aggregate indicators. The top panels of Figure 7 display the
slopes that are used by our 8-state web calculator. It seems
most sensible to forecast the aggregate associated with the
ratings you are most concerned with. The calculator will
also accept forecasts of both, using the default forecast for
speculative grade credit migration probabilities, and the
downgrade forecast for the investment grade ones. Figures
7a and 7b present a bewildering array of numbers, that
nevertheless have a pattern to them. This can be better
appreciated from the two lower panels of Figure 7, which
graph the information in the two tables, respectively. Here,
however, the sensitivities of the migration probabilities have
been lined up according to the number of rating notches
moved. So, the No change points represent the figures on
the diagonals of the tables, while, for example, the Up 2
point for BBBs gives the sensitivity of BBB-AA upgrades.
The investment grade graph shows that, when there is a 1%
increase in investment grade downgrades, for each rating
category this largely takes the form of a decline in issuers
with unchanged ratings, and an increase in those
downgraded by one notch. Moreover, the percentage that
shifts in this manner is similar across rating categories.
The other graph shows how a 1% increase in the aggregate
speculative grade default rate translates into changes in
individual migration probabilities. There is a decline in
both upgraded issuers and the no change category, which
is distributed not only into one-notch downgrades, but also,
where possible, into deeper downgrades. In contrast to the
investment grade case, the sensitivities differ among rating
categories, the shifts in migration probabilities being more
marked, the lower down the rating spectrum we go.
Table 11
Rock-bottom spreads under alternative credit

downswing views
Historical
AAA
12
months
5
5
14
36
months
120
months
16
23
82
30
92
34
102
50
143
230
260
289
316
434
506
570
623
667
797
1036
1167
1238
1277
1291
13
A
BBB
24
72
27
BB
B
CCC
24
months
15
AA
Figure 1, presented on the first page of this paper, shows

rock-bottom spreads for the actual bonds in the market,
aggregated into broad rating categories, on the assumption
that the credit downswing lasts only through the next 12
months. The figures differ from those in the second column
of Table 11, which nominally embody the same credit view.
For example, in Figure 1, the B-rated rock-bottom spread is
868bp, while the corresponding number in Table 11 is 570
bps. We can use the information presented so far,
describing the sensitivity of rock-bottom spreads to changes
in input assumptions, to try to reconcile this difference. A
summary of the effects involved is presented in Table 12.
Table 12
Reconciling BB Rock Bottom Spread Numbers

Table 11 vs. Figure 1
Generic Bond
8 states
(Table 11)
B-rated rock-bottom spread
To illustrate, say that for as long as the current credit

downswing lasts, investment grade downgrades are
expected to be 11.5% and speculative grade defaults are
expected to be 8%. Both of these figures are 4% higher than
their respective historical averages. Consequently, to
construct the investment grade rows of the credit migration
matrix for the downswing years, add four times the top four
rows of Figure 7a to Table 4. Similarly, the speculative
Downswing lasts for next
Credit
Migration
Inputs
Maturity
Coupon
Recovery rate volatility
Baseline Default Rate
High-Yield Index
Change in
18 states
Rock-Bottom
(Figure 1)
Spread
Source
570
868
298
10yrs
8%
$45/$100
$0
7%
7.3yrs
9.93%
$35/$100
$23
8.96%
30
32
100
16
95
Total
273
Table 6
Table 8
Table 10
Table 9
Table 12
New York
October 25, 2001

Portfolio Research
First, the average maturity of B-rated (senior, unsecured)

bonds in the JP Morgan High Yield index is 7.3 years,
compared with 10 years for the generic bond valued in
Table 11. From Table 6, the spread difference between 7
and 10-year B-bonds is 33bps, and so we estimate that the
effect of the 2.7 year differences is 30bps
(2.7yrs/3yrs*33bps).
Second, the average coupon of the index bonds is 9.93%,
against the 8% assumed for the generic 10 year. Table 8
shows a 129bps difference in rock-bottom spread between
8% and 16% coupon B-bonds, so we interpolate that
moving from an 8% coupon to a 9.93% coupon adds 32bps
to the rock-bottom spread.
Third, Figure 1 assumes a recovery rate of $35 per $100 of
principal, in line with current prices of bonds in the month
after they default. This is $10 lower than the baseline
recovery rate and from the last column of Table 10 we
estimate that it should raise the B-rated rock-bottom spread
by 100bps.
Fourth, the baseline recovery volatility is zero. Table 9
shows that B-rated 10-year rock-bottom spreads are 33bps
higher when recovery volatility is at its maximum of $50.
Interpolating once again, , the historical average of $23 used
in Figure 1 raises the rock-bottom spread by 16bps.
The largest effect comes from the fact that Figure 1 is based
on the default and credit migration numbers for modified
ratings (B1/B+, etc). The composition of the B-rated sector,
is different now from its past average, that resulted in the
7% average default rate in Table 4. So, when we aggregate
the historical default rates of B1, B2 and B3-rated bonds,
using current outstanding amounts, the resulting default rate
is 8.96%, because B3 and B2 bonds are now more prevalent.
Now, Figure 7 shows that a 1.53% rise in the B-rated
default rate corresponds to a 1% rise in the aggregate
speculative grade default rate. So, having a B-rated default
rate 1.96% higher than the baseline scenario corresponds to
an aggregate default rate that is 1.28% higher in each year
(1.28=1.96/1.53). Now we can use Table 11 to estimate the
spread effect of this elevated default rate. The Table shows
a 297bps difference between rock bottom spreads under the
historical credit fundamentals scenario, and under the
scenario where defaults are 4% higher for 10 years.
Consequently, our 1.28% higher speculative grade default
rate would correspond to 95bps (1.28%/4%*297bps).
Adding up all these effects gives 273bps, compared to the
298bps difference between the two B-rated rock-bottom
spreads with which we started. The remaining 25bps is the
result of second-order effects of the sources discussed
above (we only used linear interpolation), and of
interactions among these effects. Nevertheless, it is
remarkable how closely one can account for the differences
using just these basic tools.

page 27
Where to go from here

Our exhaustive tour of the mechanics of rock-bottom
spreads has, for the sake of clearer exposition, focussed
exclusively on plain vanilla bullet bonds of a generic issuer
in each rating category. The real world is somewhat more
complicated in several ways, and it is well to discuss
whether the rock-bottom spread framework is equal to the
challenges.
One complication is that actual bonds present more
elaborate patterns of cashflows than plain vanilla. If
anything, this is where the strength of the rock-bottom
spread approach lies, as it can deal with any stream of
cashflows subject to credit fundamentals risks. Callable and
puttable bonds, discount bonds, subordinated paper, Brady
bonds, structures with cashflows contingent on ratings
changes, can all be valued in terms of their rock-bottom
spread or price. Moreover, since these valuations emanate
from a single framework, it is possible to make consistent
relative value assessments across bonds with very different
cashflow structures. Because the market does not presently
have consistent yardstick for comparing value among, say,
discount bonds and subordinated bonds, callable or not, this
may be the place to look for unexploited pockets of value.
While the rock-bottom spread framework is punctilious to a
fault on cashflow structure, it uses only the limited
information on issuers credit quality contained in ratings.
The problem this creates is that bonds that appear cheap for
their rating may be expensive because of the special credit
conditions of the issuer, or vice versa.
One way around this is to look at large portfolios or
averages of bonds, rather than individual bonds. This is
obviously the appropriate route for comparing value across
broad asset classes. The cheapness or dearness of the bonds
as a group cannot be vitiated by issuer idiosyncracies,
because these have to average out by definition. For
example, if B-rated bonds are currently expensive relative
to their rock-bottom valuation, then only if todays B-rated
bonds are better quality than those in the past can one
escape the conclusion that as a group they are expensive.
This does not seem to be so at the moment.
However, where the goal is to pick individual bonds, the
rock-bottom spread valuation has to be complemented with
better issuer-specific information, either directly from credit
analysts who follow the industry or issuer in question, or
from the equity market. In this way, it is possible to qualify
the cashflow valuation with a health check on the issuer.
Bonds that appear cheap and healthy should be expected to
perform better than those that offer a positive signal in only
one, or neither of these respects .
New York
October 25, 2001

Portfolio Research
Examples of each of these uses of the rock bottom valuation

are now readily available
The effects of callability, discount structures and
subordination can be examined using the new RockBottom Spread Calculators on the Credit tab on
www.morganmarkets.com [1]. See also Valuing
Rating-Triggered Step-up Bonds [6]
References
1 Introducing the J.P. Morgan Rock-Bottom Spread
Calculator, July 12, 2001, Mansoor Sirinathsingh
2 Emerging Market Collaterlaized Bond Obligations: An
Overview, October 1996, Moody's Investors Service
3 Valuing Credit Fundamentals: Rock Bottom Spreads,
November 17, 1999, Peter Rappoport
4 Default and Recovery Rates of Corporate Bond Issuers:
2000, Moody's Investors Service
5 Almost Everything You Wanted to Know about
Recoveries on Defaulted Bonds, Edward S. Altman and
V. Kishore, Financial Analysts Journal, Nov/Dec 1996,
pp 57-64.
6 Valuing Rating-Triggered Step-up Bonds, Mansoor
Sirinathsingh
7 Introducing the Rock-Bottom Roundup, July 12, 2001,
Peter Rappoport, Mansoor Sirinathsingh
8 The Rock-Bottom Roundup, July 12, 2001, Peter
Rappoport, Mansoor Sirinathsingh
9 Picking High-Yield Bonds, July 12, 2001, Peter
Rappoport, Mansoor Sirinathsingh
10 Rock-Bottom Spread Tutorial (Excel spreadsheet)
Frank Zheng. Accessible on the Credit tab on

page 28
A comparison of market spreads and rock-bottom

spreads across asset classes is given in the Rock-Bottom
Roundup [7,8], which compares US High Grade and
High Yield Corporates, and Emerging Markets
Sovereigns
Picking High Yield Bonds [9] describes the
performance of a rule that selects bonds whose rockbottom valuation is not contradicted by their isuers
equity performance.
Portfolio Research
New York
Peter Rappoport
Mansoor Sirinathsingh
Frank Zheng
(1-212) 648-1268
(1-212) 648-4915
(1-212) 648-1860
London
Alan Cubbon
Guy Coughlan
Stephen Tang
(44-20) 7325-5953
(44-20) 7777-1857
(44-20) 7777-1534
Norway
Halvor Hoddevik
(47-22) 941-978
Tokyo
Tatsushi Kishimoto
(81-3) 5573-1521
New York
October 25, 2001

Portfolio Research
J.P. Morgan Securities Ltd.
Portfolio Research
Luca Brusadelli (44-20) 7325-5607

page 29
October 17, 2001
Portfolio Research
Valuing European Credit Fundamentals

Should we value the credit exposure of European
and US corporates in the same way?
Figure 1
Implied rates of downgrade to below BBB: US vs. Europe

US minus Europe, %
historical patterns of rating changes have been

very different
European fallen angels have been far scarcer
25.0
We trace the implications of these differences for

European issuers credit spreads
15.0
Well-established A-rated industrials offer better

value than recently-rated issuers
BBB
20.0
10.0
5.0
AA
AAA
0.0
To value any corporate bond, it is necessary to take a

view on the likelihood that its credit rating will change. A
common practice is to base this view on historical
frequencies of rating transitions by the rating agencies.
Twenty years of experience is available for a large
universe of US obligors; for European issuers, only onetenth as much information exists. Although as a result
the European experience is less statistically reliable, it
suggests markedly different credit dynamics (Figure 1).
Our analysis shows that these differences in exposure to
downgrades translate into significant differences in
value. As shown in Figure 2, BBB-rated bonds for whose
valuation the European experience is relevant can trade
some 36bp tighter than their US counterparts, and still
pay adequately for their credit exposure. A-rated European-style issuers can trade 18bp tighter.
Which issuers should be associated with which credit
fundamentals? It does not seem appropriate to make the
determination solely on the basis of domicile. Rather,
new entrants to European bond markets seem to place a
greater emphasis on delivering equity returns, and
therefore we view them as more in the US mould.
Accordingly, we value European companies that have
first come to the Eurobond market in the last few years
using the US rating transition experience. We value the
more seasoned European issuers on the basis of the
European rating transition data.
Our distinction between seasoned and recent European
issuers appears to accord with the markets valuation for
BBBs (Figure 3). Recent BBB issuers spreads exceed
years
10
Figure 2
Rock-bottom spreads on a 10-year 7% bond

asset swap spreads, bp as of 12Oct01
50
European transition
40
US transition
30
20
10
0
-10
-20
AAA
AA
BBB
Figure 3
Industrials excluding Media & Telecoms and Technology

180
160
140
120
100
80
60
40
20
0
-20
Rock-bottom
Market
A seasoned
A recent
BBB seasoned
BBB recent
seasoned spreads by about 31bp, as does our independent valuation of their credit exposure. However, there is
an anomaly among A-rated names, where established and
New York
October 25, 2001

Portfolio Research
recent issuers spreads are roughly the same, while our

analysis suggests that the seasoned issuers should trade
15bp tighter.

page 30
Figure 4
One-year rating transition rates: US vs. Europe

US minus Europe, %
4.0
AAA
Credit migration: US vs. Europe
3.0
AA
2.0
Credit exposure concerns the potential for losses from

an obligors default. Hence, probabilities of changes in
credit quality and default, as well as expected recovery
rates, are the most relevant sources of credit risk. How
can we quantify these credit fundamentals? Credit
migration probabilities can be described by the historical
frequencies of upgrade, downgrade and default by the
rating agencies. The historical experience of rating
migration is typically summarized in a one-year transition
matrix containing historical averages of the annual
frequencies of rating changes and defaults for each
rating category.
1.0
BBB
Our quantitative description of the sources of credit

exposure is driven primarily by the information contained
in credit ratings. But does the same credit rating entail
the same exposure to downgrade or default for firms in
the same industry or geographical area? Here, we look at
the differences in the rating migration experience of US
and European corporates. Thus, we compare the transition matrix for US issuers with that drawn from the
rating performance of European obligors. The two
matrices, shown in Table A1 and A2 of the Appendix, are
calculated from the annual frequencies of rating changes
by Standard & Poors over the period 1983-2000.
Figure 4 summarizes the differences between the US and
the European investment-grade transition rates in a
convenient way. For each rating, the differences in the
one-year frequencies of rating transition (upgrade/
downgrade by different notches) and rating stability are
shown. The rating transitions of AAA-, AA- and A-rated
issuers have been quite similar in the US and Europe.
What really stands out from the comparison is the
difference in the historical incidence of rating changes
for BBBs. European BBB companies have on average
enjoyed a greater rating stability and a higher upgrade
rate than their US peers. Most important, the BBB oneyear frequency of downgrade to speculative grade has
been much lower than that experienced by US issuers. In
fact, there has been only a handful of so-called fallen
angels in the history of European rating transitions:
contrary to the US experience, the BBB rating has been a
floor to the credit migrations of European investmentgrade issuers.
0.0
-1.0
-2.0
-3.0
-4.0
up 3
up 2
up 1 unchanged down 1
down 2
down 3
down 4
down 5
down 6
Each line plots the differences between the cells of the corresponding
row of the matrices contained in Table A1 and A2. For example, the
chart shows that BBB ratings in the US have been downgraded to BB
with an annual frequency around 3.5% higher than in Europe.
To emphasize the implications of these differences, we

also look at the rates of downgrade to speculative grade
(including default) at longer horizons as implied by the
one-year transition frequencies. Figure 1 (on the front
page) shows for each rating the difference in these rates
between the US and the European transition matrix. For
all ratings and at all horizons, US rates are higher than
those implied by European one-year transitions. Moreover, the gaps widen as the horizon extends. The extreme
case is that of BBBs: over ten years a European company
initially rated BBB is subject to a probability of ending in
a speculative grade category, or defaulting, 22% lower
than that of a US company.
To assess the importance of these differences in credit
fundamentals, we need to convert them into price or
spread terms. Specifically, we address the following
question: how does a bonds value change under the
alternative views that its issuers future rating performance will conform to the historical experience of either
the US or Europe?
Pricing credit exposure: rock-bottom spreads

To quantify this difference, we use JPMorgans Rockbottom Spread framework*. This translates a bonds
credit fundamentals into a reservation price (the rockbottom price): the highest price an investor should be
willing to pay for bearing the bonds credit risk, consistent with a given investment performance objective. The
* See references for further details on the Rock-bottom Spread framework
New York
October 25, 2001

Portfolio Research
to a Europeantype credit should demand a spread much

lower than required for covering exposure to a US-type
issuer of the same rating.
Table 1
Rock-bottom spreads

5-year 7% bond
US matrix
Rec.=45%
AAA
AA
A
BBB
-18
-10
-4
30
10-year 7% bond
US matrix
Rec.=45%
AAA
-16
AA
-7
A
4
BBB
43
European matrix
Rec.=45% Rec.=22.5% Rec.=0%
-19
-18
-16
8

page 31
-19
-18
-15
17
-19
-18
-14
26
European matrix
Rec.=45% Rec.=22.5% Rec.=0%
-19
-19
-19
-17
-17
-16
-14
-12
-10
7
16
25
These figures are produced with the 8-state transition matrix, using the
following inputs:
Recovery volatility: 23%
Diversity score: 57
rock-bottom spread associated with this price is the

lowest spread an investor should require for holding the
bond.
The ingredients for calculating rock-bottom spreads
include, as well as the rating transition probabilities, an
estimate of how much will be recovered from a defaulted
bond. As a baseline, we assume a recovery rate of 45 per
100 of nominal; this is the (US) historical average of
prices of bonds after default. To account for the opportunity to diversify credit risk, we refer to a portfolio
diversity score appropriate for the entire investmentgrade corporate bond market. The investment performance objective underlying our framework is based on a
target information ratio of 0.5 against an investment in denominated government bonds.
Table 1 shows the resulting rock-bottom spreads at the 5and 10-year maturity for all investment-grade ratings.
Here, rock-bottom spreads are expressed as asset swap
spreads in line with the bond market convention. What
are the implications on rock-bottom spreads of the
differences in the US and European historical transition
rates? As we would expect from the comparison of
transition frequencies, rock-bottom spreads based on
European credit migration are lower than those arising
from US transition rates. This is true for all ratings and at
both maturities. Also, the gap between spreads widens as
we move down the credit rating scale. The largest gap is
found in BBBs: European rock-bottom spreads at 5 years
fall below US spreads by 22bp; at 10 years the difference
is as much as 36bp. Hence, an investor holding exposure
To this point, we have not accounted for any potential

difference in recovery rates. What would be the effect
on rock-bottom spreads if we expected a lower recovery
on those bonds subject to European transition probabilities? We look at two scenarios: one where half of the US
historical average is recovered (22.5%) and one where
zero is recovered. The right-most columns of Table 1
show the implications on European rock-bottom spreads.
High credit-quality spreads are only marginally affected
by reduced recovery rates. The only significant effect is
on BBBs: rock-bottom spreads rise by around 9bp under
each scenario. Even in the extreme case of a zero
recovery rate, European BBB spreads are lower than US
ones: the gap is reduced to 4bp at 5 years, but still
doesnt fall below 18bp at 10 years.
Thus, even allowing for extremely low European recovery rates, the conclusion does not change: according to
historical rating experience European issuers, particularly
BBBs, have carried a much lower exposure to credit risk
than US issuers of the same rating.
It follows that any view on the European credit market
must focus on whether future rating transition rates are
expected to conform to historical averages or converge
to the US experience. Depending on this expectation,
there are sizable differences in the credit spreads investors should demand for bearing exposure to European
credits. In turn, these differences are crucial for assessing the value offered by current market spreads.
European spreads: market vs. rock-bottom

How do we expect European ratings to perform in the
future? We believe that increasing leverage by European
investment-grade companies will be the main force in the
convergence of transition rates to US averages. Its likely
that issuers driving this process will be found among
recently-rated names of lower credit quality rather than
well-established corporates. Specifically, we look at the
universe of Western European industrial names drawn
from the current composition of the JPMorgan Maggie
Investment Grade Credit Index. From this universe we
define:
recent issuers as those names that have been assigned their first rating as A or BBB in the last three or
five years, respectively;
New York
October 25, 2001

Portfolio Research
Table 2
Breakdown of European Investment Grade Industrials

As of 12Oct01
Seasoned issuers
Bonds Mkt value, bn
AAA
AA
A
BBB
3
38
94
41
1.1
30.8
81.6
41.1
Recent issuers
Bonds Mkt value, bn
29
51
17.6
33.4
Invensys, KPN and Sonera have been excluded from the analysis.
Source: JPMorgan Aggregate Index Euro (MAGGIE)
seasoned issuers as all remaining companies.

For recent issuers we assume future transition rates
equal to those experienced by US companies over the
last twenty years. Seasoned issuers are instead expected
to conform to European historical transition rates. This
partition of our universe results in a breakdown of the
index constituents as shown in Table 2.
We compare the credit exposure, as implied by the credit
rating, entailed by the two group of names. To do so, we
calculate average rock-bottom spreads per rating on the
portfolios of bonds issued by each group. We focus only
on A- and BBB-rated bonds as all recent issuers are
concentrated in these two rating categories. As shown in
Figure 5, the pattern of the resulting rock-bottom
spreads reflects the higher credit risk underlying recent
issuers: their A and BBB rock-bottom spreads exceed
those on seasoned issuers by 15bp and 34bp, respectively.
To what extent are these differences in credit exposure
priced in by the market? From Figure 5, we see that BBB
market spreads broadly reflect the difference in rockbottom spreads. The comparison looks quite different
when we consider A-rated bonds. The higher credit
exposure of the portfolio of recent A-rated issuers is
actually associated with a lower market spread. In terms
of premium over rock-bottom, market spreads on
seasoned issuers pay 30bp more than those on recent
issuers.
In our framework, the excess of market spreads over
rock-bottom is a premium over what would be required
to cover credit exposure: it compensates for the lower
liquidity of corporate versus government bonds. There is
no reason to believe that seasoned issuers are any less
liquid than recent ones. So, in terms of exposure to longterm credit fundamentals, seasoned A-rated issuers are
significantly underpriced compared to recent issuers of
the same rating.

page 32
Figure 5
European industrials: market vs. rock-bottom spreads

200
180
160
140
120
100
80
60
40
20
0
-20
Market
RBS
Excess
Rock-bottom
Market
A seasoned
97
-17
114
A recent
82
-2
84
BBB seasoned
137
-11
148
BBB recent
188
23
165
Figure 6
Excluding Media & Telecoms and Technology

200
180
160
140
120
100
80
60
40
20
0
-20
Market
RBS
Excess
Rock-bottom
Market
A seasoned
80
-17
97
A recent
83
-2
85
BBB seasoned
124
-7
131
BBB recent
155
24
131
Figure 5 and 6 show duration-weighted average rock-bottom spreads

produced with the 18-state transition matrix. The inputs used are the
same as for Table 1 with the exception of the recovery rate for which
38% is assumed, in line with current traded prices of defaulted debt.
The results in Figure 5 are partially biased by some

sector-specific factors. If we exclude issuers in Media &
Telecoms and Technology, whose ratings have a high
degree of uncertainty, market spreads on all portfolios,
except that of recent A-rated issuers, are significantly
reduced as shown in Figure 6. In terms of excess over
rock-bottom spreads, we do not find any relative-value
opportunity in BBBs: market spreads exactly compensate
for the difference in credit exposure of the two portfolios. While the average market spread on seasoned Arated issuers drops substantially by 17bp this portfolio still pays in excess of rock-bottom 12bp more than
that of recent issuers.
New York
October 25, 2001

Portfolio Research
Conclusion
The experience of rating transition has differed significantly in the US and Europe. The assessment of credit
exposure to European corporates depends critically on
which pattern of transition we view as more appropriate
for the future. Consequently, sizable relative-value
opportunities may arise. Our analysis suggests that wellestablished A-rated industrials offer better value than
recently-rated issuers.

page 33
Appendix
Table A1
US average one-year transition rates

percentage
AAA
AA
A
BBB
BB
B
CCC
AAA
AA
A BBB
BB
B CCC
D
93.63 5.88 0.37 0.09 0.03 0.00 0.00 0.00
0.65 91.72 6.96 0.51 0.02 0.10 0.02 0.01
0.08 2.09 92.10 4.99 0.50 0.21 0.01 0.03
0.03 0.27 4.89 89.32 4.28 0.84 0.12 0.24
0.03 0.05 0.42 6.81 84.08 6.58 1.01 1.03
0.00 0.11 0.30 0.46 5.81 83.69 3.57 6.07
0.17 0.00 0.34 1.02 2.05 11.26 59.04 26.11
Table A2
European average one-year transition rates

percentage
AAA
AA
A
BBB
BB
B
CCC
AAA
AA
A BBB
BB
B CCC
D
92.99 6.72 0.29 0.00 0.00 0.00 0.00 0.00
0.36 91.48 7.73 0.43 0.00 0.00 0.00 0.00
0.00 2.37 93.15 4.18 0.30 0.00 0.00 0.00
0.00 0.19 6.32 92.34 0.77 0.19 0.00 0.19
0.00 0.00 1.07 4.81 87.70 6.42 0.00 0.00
0.00 0.00 1.56 1.56 4.69 82.81 4.69 4.69
0.00 0.00 0.00 0.00 0.00 0.00 75.00 25.00
Source: JPMorgan, based on Standard & Poors database of credit

rating history over the period 1983-2000
New York
October 25, 2001

Portfolio Research

page 34
New York
October 25, 2001

Portfolio Research
Mansoor Sirinathsingh (1-212) 834-9224
Portfolio Research

page 35
September 10, 2001
Using the JPMorgan Rock-bottom

Spread Calculators
Web-based tools implementing JPMorgans proprietary credit valuation framework
Allows uploading of client portfolios as well as the
use of the JPMorgan High Yield Index
Permits sensitivity analysis
The JPMorgan Rock-bottom Spread calculators are

web-based tools implementing our proprietary credit
valuation framework. Basically, a rock-bottom spread
(RBS) is the minimum spread you need to be paid on a
bond, or a group of bonds, to earn sufficient return for

the credit risk involved (see Valuing Credit Fundamentals, November 1999).
All clients with a Morganmarkets ID automatically have
access. Follow the links on the Fixed Income or Credit
pages. These links should take you to the main page of
the calculators. From this screen, shown in the first
figure below, there are three calculator options. The first
two cover U.S. corporates and differ in the granularity
of the bond ratings. The third covers Emerging Markets
sovereign debt. The following figures demonstrate how
to use these calculators, and highlights the key inputs.
Launches 8-state calculator
New York
October 25, 2001

Portfolio Research
If only this box is checked, all

elements (investment and
speculative grades alike) of
the transition matrices for
each future year will be
adjusted in proportion to your
default forecasts.

page 36

forecast of investment grade
downgrades.
If both boxes are checked,

investment grade elements
of the transition matrices for
forecast of investment grade,
downgrades, and the
speculative grade elements
forecast of speculative grade
defaults.
Create your forecast by

dragging red points to
appropriate levels.
New York
October 25, 2001

Portfolio Research

page 37
8-state calculator
View the transition
matrix associated
with each year of
your forecast.
Complete your credit

fundamental forecast by
entering a recovery rate
and volatility.
Use either a flat

government curve by
entering rates or the
forward curve from
paper Rock Bottom
spread mechanics.
The settings on this line

override the settings on
the lines below and allow
you to set default bond
characteristics.
Enter the relevant

information for the
bonds you wish to
price.
New York
October 25, 2001

Portfolio Research

page 38
18-state calculator
Launches 18-state
calculator. This prompt will
initially lead to same page
as 8-state calculator

default forecasts.

forecast of investment grade
downgrades.
If both boxes are checked,

investment grade elements
forecast of investment grade,
downgrades, and the
speculative grade elements
forecast of speculative grade
defaults.
Create your forecast by

dragging red points to
appropriate levels.
New York
October 25, 2001

Portfolio Research

page 39
18-state calculator
Download file
template, enter
portfolio bonds
and save
Select saved file
These extra options in the

18-state calculator enable you
to upload your own portfolio or
value the JPMorgan HY index.
Determine price cutoff such that

index bonds with lower prices are
not included in the valuation.
Select a particular sector to

value or use All bonds to
value the entire index.
New York
October 25, 2001

Portfolio Research

page 40
18-state calculator
Summary of bond
characteristics.
Rock Bottom
spreads expressed
in basis points.
Run sensitivity
analyses for
each bond.
Summarizes credit
fundamentals
forecast.
Select two attributes on

which to run the analysis.
Enter ranges and step

size for sensitivity
analysis.
New York
October 25, 2001

Portfolio Research

page 41
Emerging markets calculator
Launches Emerging
Markets calculator
Set ratings and

outlooks for each
Emerging Markets
country.
To complete credit
fundamentals
information, enter
a recovery rate
and volatility.
New York
October 25, 2001

Portfolio Research

page 42
Individual bond
Rock Bottom spreads
Scroll down for

country spreads
New York
October 25, 2001

Portfolio Research

page 43
August 1, 2001
Portfolio Research
Introducing the Rock-bottom Roundup

Our new publication, the Rock-bottom Roundup
compares value among
US investment grade corporate bonds
US speculative grade corporate bonds
Emerging markets sovereign bonds
It uses JP Morgans proprietary Rock-bottom

Spread (RBS) framework to value credit risk
consistently across markets
The more the market spread of an asset class
exceeds its RBS, the better the value, all other
things equal
This note is intended as a users guide to our Rock-bottom
Roundup publication, which provides a uniform
comparison of value among US Investment Grade and
Speculative Grade Corporate bonds, and Emerging Markets
sovereign bonds. The Roundup will be published every
three months, and occasionally more frequently if changes
in credit conditions warrant.
Comparing the raw market spreads of the three asset
classes does not make sense, because their exposures to
credit and liquidity risk are so different. JP Morgans rockbottom spread framework effectively filters out the credit
risk component.
The rock-bottom spread of a bond or group of bonds is
what you need to be paid to earn sufficient return for the
credit risk involved. So, the difference between market
spread and rock bottom, or surplus spread, indicates your
reward for bearing non-credit exposures, notably illiquidity.
All other things equal, the higher the surplus spread, the
better. Of course, all other things are not necessarily
equal; an asset class with a high surplus over rock-bottom
may be bad value, because its illiquidity is too great, or its
market price volatility is too high. Looking at rock-bottom
spreads provides the right place to start to investigate these
issues.
A bonds rock-bottom spread combines the broad aspects

of its credit exposure those driven by its rating, coupon
and maturity, seniority, call provisions -- with views on
aggregate credit trends and expected recovery rates, into a
measure of how much spread is needed for bearing the
credit exposure. It does not factor in issuer-specific
conditions, such as whether their earnings are on track, or
if a patent application has been granted, except insofar as
these find their way into the issuers rating. Consequently,
it is quite plausible that there are special circumstances that
make a specific issuer good value, even though its market
spread is less than its rock-bottom spread. However, in
the aggregate, the special circumstances cannot dominate.
Thus, our emphasis is on relative value among broad asset
classes, distinguished by the characteristics that go into
rock-bottom spreads. We compare surplus spreads across
corporates and emerging markets sovereigns, and among
rating categories within each. Similarly, we compare
groups of bonds according to their seniority, and by their
cashflow structure.
Each asset class rock-bottom spread aggregates the rockbottom spreads of its component bonds, weighted by
outstanding amounts. The bond-level rock-bottom spreads
depend on three distinct types of inputs
Bond parameters:
coupon,
maturity,
seniority,
cashflow structure (bullet, discount, callable, etc),
rating
Market structure parameters:
portfolio diversification
Credit fundamentals views :
Expected pattern of future aggregate defaults and
downgrades
Expected recovery rates, and their volatility
We focus on portfolios that achieve the maximum

diversification possible in each market. Consequently,
investors can differ only on the third group of inputs. The
New York
October 25, 2001

Portfolio Research
Rock-bottom Roundup uses credit fundamentals views that

appear to represent current views in the market. In
particular, we reference the 12-month forecast of default
rates published in Moodys Monthly Default Report. We
use the most recent average of market prices of defaulted
bonds, also to be found in Moodys report, to characterize
expected recovery rates. Of course these views are only
representative. Should your views differ from the ones in
the Roundup, you can produce your very own RockBottom Roundup numbers, using the Rock-bottom Spread
calculators on the JP Morgan website.
Table 1 of the Roundup presents a cross-market
comparison for speculative grade and investment grade
corporates, and emerging markets sovereigns. For
emerging markets the spreads presented are stripped
spreads. As is consistent with market practice, we define
bonds trading at prices below 50 as distressed. Treating
these bonds as though they are certain to default, and
reducing the default probabilities on the remaining bonds
proportionately, we calculate the rock-bottom and market

page 44
spreads of the non-distressed bonds. For emerging

markets sovereigns we consider two possible scenarios
with respect to portfolio composition. The first is for the
dedicated emerging markets investor, whose portfolio
comprises only emerging markets sovereign bonds
(Dedicated). The second is for the investor who buys
emerging markets sovereigns into a maximally diversified
corporates portfolio (Crossover).
Tables 2 through 5 present for senior unsecured,
subordinated, senior secured and discount bonds, the rockbottom and market spreads, aggregated by rating category.
Given their different levels in the capital structure, each of
these types of bonds have different recovery rates, and
their rock-bottom spreads are calculated accordingly.
Tables 6 and 7 summarize the results for emerging markets
sovereigns, showing both stripped and blended spreads.
New York
October 25, 2001

Portfolio Research

page 45
Published:
Author:
Portfolio Research
August 30, 2001

Rock-bottom Roundup
B-rated U.S. corporates have not adequately priced in the recession
while BB-rated corporates offer better value
Emerging markets are particularly attractive to crossover investors
For the crossover investor, high-grade emerging markets are priced similarly to A- and BBB-rated U.S.
corporates
Table 1
Cross-market summary
Rock bottom spread
Market Spread
Difference
274
989
844
412
744
702
138
-245
-142
509
228
407
642
839
255
487
881
330
27
80
239
401
145
274
457
839
255
487
881
438
110
213
424
67
44
98
191
180
233
124
136
135
High Yield Corporates

BB
B
B Non-distressed
EMBIG (Dedicated)
Inv Grade
BB
B
EMBIG (Crossover)
Inv Grade
BB
B
High Grade Corporates

AA
A
BBB
New York
October 25, 2001

Portfolio Research

page 46
U.S. Corporates
Table 2
Senior unsecured bonds
Aaa
Aa1
Aa2
Aa3
A1
A2
A3
Baa1
Baa2
Baa3
Ba1
Ba2
Ba3
B1
B2
B3
RBS
Market
Surplus
Coupon
Maturity
# Bonds
10
8
24
26
37
30
48
67
81
135
187
270
461
663
965
1429
93
106
127
120
143
161
197
206
213
273
339
479
480
638
645
1021
83
98
103
94
106
131
149
139
132
138
152
209
19
-25
-320
-408
6.72
6.51
6.91
6.82
7.2
7.2
7.51
7.48
7.54
7.98
8.25
8.66
8.96
9.72
9.84
10.36
13.57
11.34
16.05
12.65
18.05
14.8
16.94
17.89
14.94
15.57
8.23
7.15
7.43
6.71
7.62
7.45
144
35
130
232
222
485
303
391
329
334
90
77
77
65
89
94
RBS
Market
Surplus
Coupon
Maturity
# Bonds
8
4
8
18
28
28
28
35
67
61
166
213
283
382
611
858
122
135
139
144
147
153
166
187
210
224
344
413
413
449
525
771
114
131
131
126
119
125
138
152
143
163
178
200
130
67
-86
-87
7.3
6.86
7.55
7.39
7.09
7.16
7.44
8.03
8.28
7.79
9.23
9.43
9.3
9.59
9.65
10.42
15.72
10.55
9
13.02
10.54
14.03
10.91
7.68
11.59
4.63
9.04
8.99
7.45
7.5
7.74
7.32
11
4
1
72
133
93
28
3
8
6
19
18
47
36
126
170
Table 3
Subordinated bonds
Aaa
Aa1
Aa2
Aa3
A1
A2
A3
Baa1
Baa2
Baa3
Ba1
Ba2
Ba3
B1
B2
B3
New York
October 25, 2001

Portfolio Research

page 47
Table 4
Senior secured bonds
Ba1
Ba2
Ba3
B1
B2
B3
RBS
Market
Surplus
Coupon
Maturity
# Bonds
143
238
368
463
882
1152
313
524
492
529
739
1135
170
286
124
66
-143
-17
8.41
11.09
9.77
8.75
9.88
11.24
8.89
8.08
6.89
11.01
3.73
6.79
5
2
15
10
11
12
RBS
Market
Surplus
Coupon
Maturity
# Bonds
438
--726
995
1493
443
--958
841
1337
5
--232
-154
-156
10.27
--10.55
10.77
11.75
5.26
--6.32
8.16
7.19
2
--5
11
27
Table 5
Discount bonds
Ba1
Ba2
Ba3
B1
B2
B3
U.S. Corporates forecasts used in Tables 1 - 5

IG downgrade rates: 12% for the next year,
7% thereafter
14%
12%
HY grade default rates: 9% for the next year,

4% thereafter
10%
8%
6%
4%
HY Default Rate
2%
IG downgrade rate
Recovery rates:
Senior secured:
Senior unsecured:
Subordinated:
Discounts
53%
36%
16%
36%
Recovery volatility:
Diversity score:
23%
70
0%
Source: Moodys Investors Service Default Report

2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
New York
October 25, 2001

Portfolio Research
Emerging markets forecasts used in Tables 6 - 7
Emerging Markets
Recovery rate:
Recovery volatility:
Diversity score:
Table 6
Emerging markets summary

Blended
RBS
Blended
Market
Stripped
RBS
Stripped
Market
207
407
608
228
482
825
228
407
642
255
487
881
IG
BB
B

page 48
Table 7
Country summaries
Country
Algeria
Argentina
Brazil
Bulgaria
Chile
China
Colombia
Cote d'Ivoire
Croatia
Ecuador
Egypt
Hungary
Korea
Lebanon
Malaysia
Mexico
Morocco
Nigeria
Pakistan
Panama
Peru
Philippines
Poland
Russia
South Africa
Thailand
Turkey
Ukraine
Uruguay
Venezuela
S&P
rating
CCC
B?
BB- ?
B+ ?
ABBB
BB ?
CCC
BBBCCC+ ?
BBB- ?
ABBB ?
B+ ?
BBB
BB+ ?
BB
CCC
BBB+
BBBB+ ?
BBB+
B
BBBBBBB?
CCC
BBBB
Moodys Blended Blended Stripped

RBS
Market
RBS
rating
Caa2
1266
707
1266
Caa1 ?
994
1574
1044
B1
575
859
609
B2
415
524
529
Baa1
129
188
129
A3
165
132
165
Ba2
444
440
444
Caa2
857
2131
857
Baa3 ?
285
212
285
Caa2
963
1447
963
Ba1
352
365
352
A3
93
46
93
Baa2
146
149
146
B2 ?
813
529
813
Baa2
184
211
184
Baa3
229
276
270
Ba1
363
472
363
Caa2
692
904
1113
Caa1 ? 1251
1353
1251
Ba1
328
397
328
Ba3 ?
529
616
529
Ba1 ?
423
570
423
Baa1
134
200
144
B3 ?
661
827
661
Baa3 ?
244
276
244
Baa3
238
166
238
B1 ?
819
984
819
Caa1 ? 1363
1550
1363
Baa3
252
282
252
B2
525
776
623
?? Positive/Negative outlooks
Stripped
Market
690
1652
924
658
183
126
458
2313
199
1481
369
41
143
529
209
331
451
1474
1338
400
618
575
210
831
273
164
994
1534
286
933
17.5%
7.5%
9
New York
October 25, 2001

Portfolio Research
page 49
Portfolio Research
August 22, 2001
Picking High Yield bonds

Valuation Rules
We describe a simple bond selection rule based on

each bonds rock-bottom price
its issuers recent equity performance
The rule outperformed our High Yield Index by
6% annually over the last two years
Its simplicity and success underline the usefulness, for high yield managers, of
quantitative credit valuation
equity market signals
In this paper we present a simple rule for buying and
selling high yield bonds. A manager who followed this
rule over the past two years would have outperformed
the JPMorgan High Yield Index by an average 6% per
year, while making a normal amount of transactions.
Figure 1
Annualised Returns: 6-month holding periods

Bondpicking
Rule
20%
Index
10%
0%
-10%
-20%
1999.II
2000.I
2000.II
2001.I
Average
A bonds value derives both from its cashflow features

coupon, maturity, seniorityand from the specific
fortunes of its issuer. Any attempt at valuation that
ignores either is likely to come to grief. However, it may
not be necessary to go into enormous detail on both,
especially if the goal is to outperform a benchmark,
rather than to pick the best-performing bond.
Our rule combines a precise valuation of each bonds
cashflow features with a generic check on its issuers
health. It is purely quantitative in nature making it

feasible to trawl through thousands of bonds, and value
each in a common framework.
We assess cheapness or dearness by comparing each
bonds current market price with its rock-bottom price1 ,
which is based on the issuers credit rating. This signal
can be misleading if the credit rating lags information
that market prices already reflect. So, we check the
issuers recent equity returns, which will reflect this
information. Thus, the rule comes down to the idea that,
if the bond looks cheap based on its rating, and the
equity market provides no reason to suspect the rating,
then it probably is cheap. Similarly, it recommends
selling expensive bonds whose equity has
underperformed.
Any retrospective analysis of a rule such as this runs the
risk of datamining, consciously or unconsciously. To
minimize this risk, we take a somewhat obsessive
approach, presenting our results in the order our thought
process developed. Accordingly, most of the paper
discusses the performance over the last two years of the
simpleminded rule we first thought of. This filters bonds
very crudely depending on whether they score above or
below the average of each indicator rock-bottom
prices and equity returns. Other cutoffs quartiles or
deciles, instead of averages may perform better, but
we did not pursue this. After all, our main message is not
that your cutoff should be at the quartiles or deciles of
the indicators, but rather that it has been profitable to
pay attention to these indicators.
On examining our results, we were initially surprised by
the stark contrast between the performance of the rules
Buy and Sell recommendations. The success rate of
the Buys is so high as to be statistically all but impossible to dismiss as throwing darts. In contrast, the Sells
perform quite atrociously. Yet, why should the signal of
a cheap equity outperformer be better than that of an
expensive underperformer? On further reflection, we
1 This proprietary JP Morgan valuation framework provides a
price for any set of cashflows subject to credit risk. See the
references at the end of the paper for further information
New York
October 25, 2001

Portfolio Research
trace this to the intrinsic nature of bonds (versus equity)

and the way in which news about issuers affects them.
In essence, bad news typically relates to specific events,
while good news for bonds tends to be the absence of
bad news. Thus, bad news will be reported, and tend to
find its way quickly into prices, while good news is
more diffused, and may leave small amounts of money
on the table, to be picked up by a rigorous relative
valuation discipline.
A rule that just sells extreme equity underperformers,
irrespective of their rock-bottom valuation, turns out to
perform very well. However, this raises a further question: why is it that information already priced into the
equity market for six months and therefore quite stale
by the time our rule sees it is valuable in picking highyield bonds?
We initially focused on the most recent two years,
because reliable equity data were not available earlier.
Subsequently, we were able to lay our hands on earlier
equity returns, and the Postscript to the paper records
the results of the basic rule since 1996. Interestingly, the
rule performs well from the second half of 1998 onwards, but barely outperforms prior to that date. Indeed,
during the bull market prior to mid-1998, individual bond
returns seemed to hug the index return very closely. As
usual, a rising tide allows very little discrimination on
the basis of fundamental value. Thus, the rules real
value appears during a bear market. Of course, this is
when anyone with an aversion to risk would most want
a strategy to outperform their benchmark.
Last, our rule does not boil down to any more familiar
strategy. Its recommendations are spread across industry
sectors and maturities, and it does not correspond to
buying bonds that are simply low-priced relative to par.
Nor does it favor large or small issue sizes. Our view is
that it makes money because it applies a uniform quantitative valuation to a market where strategy focuses
predominantly on qualitative information about the
health of individual issuers and bonds. Outperformance
is, of course, a quantitative matter.
The bond-picking rule
All other things equal, a bond is more attractive the
higher its promised cashflows, the lower its likelihood of
going into default, the greater its seniority, the more
remote its chance of being called, and so on. The bonds
rock-bottom price explicitly accounts for these features,
and can be compared with its market price to assess

page 50
whether its credit exposure is cheap or dear. However,

the rock-bottom valuation describes the issuer only in
terms of its credit rating. Market prices may differ from
rock-bottom, not because they value the bond incorrectly, but because they reflect more timely or more
detailed information about the issuer. So it is necessary
to enforce the all-other-things-equal clause strenuously.
Our rule does this by first passing the bonds in the JP
Morgan High Yield Index through two filters:
1.Vanilla filter We first remove distressed bonds,
(those trading below $50), triple-C rated bonds and
unrated bonds. Issuer-specific information bulks large
(relative to what we know about cashflows) here, so
our approach is unlikely to offer any advantages.
2.Equity filter We use equity prices as a timely indicator of issuers health, and so we remove bonds whose
issuers equity has not traded, or does not exist. For
these bonds, we cannot assess whether all other things
are equal.
We then build an index of the equity returns of the
remaining issuers over the preceding six months, which
we use to assess each remaining bonds equity signal:
Positive equity signals: the issuers equity return over
the preceding six months exceeded the equity index
return;
Negative equity signals: the issuers equity return over
the preceding six months fell below the equity index
return;
With issuer-specific conditions summarized in this way,
we proceed to value the cashflows of each bond that
passed the vanilla filter, by calculating a credit ratio,
defined as
credit ratio =
rock bottom price

market price
We use the credit ratio to partition bonds into two

groups, according to their rock-bottom signal
Cheap bonds: whose credit ratios fall below the
average (over all the bonds)
Expensive bonds: whose credit ratios exceed the
average
New York
October 25, 2001

Portfolio Research
The rock-bottom price places a value on the pure credit

exposure of the bond, given its rating (and under a given
view of future credit fundamentals)2. These assessments
will only be reliable if the current rating is a correct
assessment of the issuers health. We have greater
assurance that apparently cheap bonds are cheap, if the
issuer also has a positive equity signal, because then
there is no reason to suspect that the rating will be
lowered. Similarly, apparently expensive bonds of issuers
with negative equity signals are more likely to be expensive, because there is less likelihood of an imminent
upgrade. This reasoning leads us to the final recommendations of the basic rule:

page 51
Figure 2
Bond selection process

All Inde x
Bonds
1598
1079
The numbers in Figure 2 denote the average number of

bonds, over the four holding periods, that found their
way into the respective category. For example, at each
trade date, equity return data were available for an
average of 431 bonds in the index. Of these an average
of 45 ended up as Buys, 143 as Sells, while for the
remaining 243, equity performance and credit ratios
provided conflicting signals.
Table 1 shows the rules decisions for a sample of bonds
on July 1, 1999, the start of our first investment period.
The letters in the first column denote the routes in Figure
2 along which the bonds were sent. Thus, for example,
the J.Crew bond was diverted into the No View
category by the vanilla filter, while the AMD bond ended
up with the same designation, but because its equity and
rock-bottom signals conflicted. While at 1.3 (versus an
average of 1.03), its credit ratio suggested it was cheap,
the 37% decline in its equity price warned that it might
be cheap for the wrong reasons.
2 For the credit fundamentals forecast, we use historical default
and downgrade probabilities and historical recovery rates and
recovery volatilities. Different credit fundamentals forecasts
shift rock-bottom prices by approximately the same amount for
bonds in similar rating categories. Since we are interested in
relative value, the actual credit fundamentals forecast prevailing
at the time of selection is not critical, and thus we simply use the
historical values
No
equity
prices
Equity
Filter
519
648
431
Buys: Cheap bonds with positive equity signals

Sells: Expensive bonds with negative equity signals
No View: bonds with conflicting signals (cheap
bonds with negative equity signals, and expensive
bonds with positive equity signals).
The way the rule partitions bonds is shown in Figure 2.
We used this rule to partition bonds into Buys and Sells
on the four dates that marked the start of each six-month
holding period (see Figure 3).
CCC,
Not Rated,
Distressed,..
.
Vanilla
Filter
Equity and
Rock Bottom
Signals
Conflicting
signals
C 243
No
Vie w
Equity: Negative
RB : Expensive
D 143
Sells
45
Buys
Equity:Positive
RB : Cheap
Table 1
Illustrative bonds from selection period 1999.II
A
B
C
C
D
E
Bond
J. Crew 10.375% 07
Regal Cinemas 9.5%, 08
Fresh Foods 10.75, 06
AMD, 11% 03
Chiquita Brands, 10%, 09
Harrahs Operating, 7.875%, 05
Market Average
Equity
Rating Return
Caa1
B2
B2
75%
Ba3
-37%
B1
-21%
Ba2
32%
18%
RockMarket bottom Credit

Price Price Ratio
98.0
94.0
100.0
82.4
0.82
101.0
130.8
1.30
100.5
97.0
0.96
96.8
105.6
1.09
1.03
Baseline Results
We now present results on the rules performance over
the four consecutive six month holding periods. At the
start of each period, we assume that the investors
portfolio is identical to the JP Morgan High Yield Index.
Recommended bonds are then bought and sold. The
volume of purchases each period is set at 30% of the
value of the portfolio (60% annually). This is on the low
side of average turnover estimates for High Yield fund
managers. The purchases are financed first from liquidation of recommended sells, and then, if necessary,
from the remaining bonds in the portfolio, in proportion
to their market capitalization. At the end of the six-month
holding period, the investor unwinds all transactions,
returning to the neutral index portfolio. To calculate
returns, we use the bid prices on which the High Yield
Index is based, subtracting a standard bid-offer spread
($1 per $100 face) for each of the two round-trips of
New York
October 25, 2001

Portfolio Research
transactions involved. Figure 3 lays out the timeline of this

process.
Table 2 provides information on the returns of the entire
index and baseline strategy. For the rest of this research
note, we will compare strategies only to the subindex of
bonds that pass the vanilla filter each period, on which we
are able to calculate meaningful rock-bottom valuations.
As it so happened, this index, whose return is shown in
the second row, was less volatile and returned more than
the full index. The baseline portfolio strategy of financing
a 30% investment in the Buy recommendations from Sells
and No View bonds, outperformed this vanilla index by
5.9% on average.
Table 2 also decomposes the investment results into Buys
and Sells, as well as showing the separate contribution of
the equity filter. Evidently, most of the outperformance of
the baseline strategy comes from the Buy recommendations. The return of the Sells actually exceeds the index in
a couple of periods, indicating that they would lose money
for the portfolio strategy, relative to the index. Sells
actually perform better without the rock-bottom filter
(-3.1% versus 1.0%), although the return of the Buys
falls by over half when the equity filter is used alone
(19.6% versus 8.8%).

page 52
Figure 3
Investment strategy timeline

Comp are
market and
rock-bottom
p rices
6 months of equity returns
Holding
period
1999.II
2000.I
2000.II
2001.I
Unwind
trades
on...
Dec 31, 1999
June 30, 2000
Dec 31, 2000
June 30, 2001
Annual returns: Basic Rule

High Yield Index
Vanilla Index
Portfolio Strategy
Buys
Equity Rule
Equity & RB
Sells
Equity Rule
Equity & RB
Table 3
Why do our Buys perform so well?

We now document that our Buy recommendations do not
boil down to some other, more familiar strategy. One
possibility is that we are carrying out sector selection
under another name. That is, does our rule outperform
because it picks bonds in industries that outperform? The
performance attribution figures in Table 4 suggest not.
Industry selection: how much we would have outperformed the index had we built a portfolio with the
Trade on
recommendations
on...
July 1, 1999
Jan 2, 2000
July 1, 2000
Jan 2, 2001
Table 2
Since it selects bonds irrespective of their size, more

precise information on how the rule works and where its
strengths and weaknesses lie comes from looking at its
performance on a headcount basis (Table 3). The success
rate of the buys is dramatically high. The probability that
our average success rate of 91% is due to luck, (given
that randomly selecting bonds as Buys would pick
outperformers 73% of the time), is 1 in 70 million. In
contrast, there is nothing remarkable about the success
rate for Sells. And this goes for the equity signal, as well
as the combined equity and rock-bottom signal.
Here, we decompose our overall performance on buy

recommendations into two components:
6 month holding period
1999.II
0.8
-0.8
1.4
2000.I
-2.3
0.8
3.0
2000.II
-9.1
-4.5
2.5
2001.I Average
9.1
-0.5
7.1
0.7
18.4
6.3
4.5
6.8
3.4
8.0
9.5
17.7
17.9
46.0
8.8
19.6
1.9
1.1
0.5
0.5
-15.3
-6.3
0.5
8.6
-3.1
1.0
Success rates for the basic rule

(% of total number of bonds)
1999.II
2000.I
2000.II
72
69
69
81
73
Buys
Equity Rule
Equity & RB
90
90
80
82
90
94
93
96
88
91
Underperformers
Index Bonds
28
31
31
19
27
Sells
Equity Rule
Equity & RB
28
25
29
33
37
27
22
17
29
26
Outperformers
Index Bonds
2001.I Average
Table 4
Industry performance attribution of buys

(% annualized)
1999.II
Index
-0.8
Industry selection
1.2
Bond selection
6.3
Buy recommendations 6.8
2000.I
0.8
2.0
5.3
8.0
2000.II
-4.5
10.3
12.0
17.7
2001.I
7.1
16.3
22.6
46.0
New York
October 25, 2001

Portfolio Research
industry breakdown of our buy picks, but with the

same bonds within each industry as the index.

page 53
Figure 4
Outperformance by Industry: Buy recommendations
Bond selection: Using the Buy rules weighting of each

industry, this aggregates the Buy rules outperformance
of the index in each industry. So, if the Buy rule selects
no bonds in a particular industry, that industry contributes zero to bond selection.
U TILITY
TRANSP OR TA TIO N
S ERV ICE
RE TA IL
M E TA LS
M INE RA LS
In each holding period, the contribution of bond selection

is significant, and typically exceeds that of industry
selection, by a significant margin. However, it is worth
noting that the industry selection aspect of the buy rule
performs quite handily as well.
M E D IA
ENTE RTAINM ENT
M A NU F ACTURING
INF O RM ATIO N
TECH NO LO G Y
H O U S ING
Figure 4 plots outperformance by industry. The Figure

shows 20 industries for each of the four six month
periods. In all but 15 of these 80 possibilities, the strategy
outperformed. This is consistent with the notion that our
picks perform well, irrespective of their industry.
H E ALTH CARE
G A M ING /LEIS U RE
FO REST PROD
CO NTAINE RS
FO O D/TO BACCO
A similar analysis shows that our buy recommendations

do not reduce to advocating buying certain maturities.
The maturity profile of the portfolio of buys is very close
to that of the vanilla index. Likewise, neither large or
small issue sizes are represented among the Buys differently from the index as a whole.
Another possibility is that the rock-bottom filters designation of cheap and dear bonds amounts to nothing more
than recommending the purchase of bonds that are cheap
in each rating/maturity bucket (for example, Ba1s with 57 years until maturity ). After all, once the rating and
maturity have been controlled for, there should only be
minor differences in rock-bottom prices between bonds.
To examine this possibility, we look at the performance
of bonds whose equity outperformed in the preceding six
months, and whose market prices are below the average
of their rating/maturity subsector. So there is no rockbottom calculation involved here at all.
A low-priced bond is defined as one whose market price
is below the average of the rating/maturity bucket it
occupies. Rating/maturity buckets were constructed
from the six ratings: Ba1, Ba2, Ba3, B1, B2, B3, and the
five maturity buckets: 1-3yrs, 3-5, 5-7, 7-10, 10+.
Table 5 aggregates the performance of this rule over
thirty categories (6 rating categories * 5 maturity buckets). The rule is inferior to the basic Buy rule, which uses
rock-bottom prices, and even underperforms the index in
FO O D A ND D RUG
FINANCIAL
ENER G Y
CO NS U M ER
NO N-D U RABLES
CO NS U M ER
D U RABLES
CH EM ICALS
AE ROSP ACE
% return
-1 0%
0%
10 %
1 9 99 .II
2 00 0.I
20 %
20 0 0.II
30%
40 %
20 01 .I
Each bar represents the excess return of the basic rule portfolio
over the index portfolio. If the rule picked no bonds in an industry,
then for our buy picks it is assigned a zero return.
Table 5
Annual returns (%): Buy Low-priced Bonds rule

Vanilla Index
Basic Buy Rule
Buy Cheap Bonds
1999.II
-0.8
6.8
7.0
2000.I
0.8
8.0
0.6
2000.II
-4.5
17.8
10.2
2001.I Average
7.1
0.7
46.0
19.7
27.8
11.4
2000I. This latter gap is of considerable interest, because

it shows that the rock-bottom rule outperforms not only
by selecting cheap bonds within the rating maturity
buckets, but also by incorporating a valuation between
these buckets. Just buying cheap bonds has nothing to
offer in this respect.
New York
October 25, 2001

Portfolio Research
Our Buy recommendations thus seem to identify a

source of value in the high-yield market that does not boil
down to any of the other obvious suspects. This lends
support to the perspective on the market we suggested at
the outset: if a bonds looks cheap according to a rigorous comprehensive valuation, and the equity market
provides no reason to smell a rat, then the bond probably
is a good deal.
Why are our Sells such a disaster?
As we have seen, the basic rule is not symmetric.
Expensive bonds whose equity has underperformed do
not lose money with anything like the same reliability.
This is partly a reflection of the crudeness of the rule we
have set up: just because an issuers equity has not risen
as much as, or fallen more than, the average, it does not
mean that its ability to meet bond obligations has been
affected materially. The issuer may still have a large
cushion of equity left even after a sizeable fall in market
value.
Indeed, lowering the threshold below which equity
returns signal a sell dramatically improves the performance of the equity filter. Table 6 shows that merely
selling bonds whose equity has fallen more than 30% in
the preceding six months would have produced an
average annual return of 17.7%. This combines with an
average success rate of 48%, which, as with the basic
Buy rule, is off the statistical significance charts.
One would think that using the extra information provided by the rock-bottom signal would improve this
outcome, but the result is an unmitigated disaster,
producing markedly worse returns and success rates in
three of the four periods. Moreover, if we make the
threshold for the rock-bottom signal more stringent, so
that less bonds are designated as expensive, matters get
even worse. For example, in 2001.I, of the 14 most
expensive bonds whose equity fell more than 30%, only
two experienced returns below the index. This result is
so extreme that there is little more reason to suspect it to
be a chance outcome than there is to believe that our
high success rate on Buys is the result of chance. So, we
need to understand how a rule designed to single out
expensive bonds systematically manages to find those
that appreciate.
The first clue is that 10 of the 14 the bonds picked by
the rock-bottom filter have prices of 92 or more. One
may wonder how this can be consistent with a 30%
decline in equity prices. Frequently, further investigation
into the specific situation of the issuer reveals a large

page 54
Table 6
Revising the negative equity signal: 30% + decline

1999.II
Vanilla Index
Sells
Equity Rule
Equity & RB
Vanilla Index
Sells
Equity Rule
Equity & RB
2000.I 2000.II 2001.I Average

Annualized returns (%)
-0.8
0.8
-4.5
7.1
0.7
-11.1
-5.2
-5.7
-6.5
-39.8
-17.7
-14.2
-3.9
% of bonds underperforming index

28
31
31
19
48
36
42
55
61
44
40
32
-17.7
-8.3
27
48
42
equity cushion, which immunizes the bonds against the

decline. Similarly, the prices of the remaining bonds with
negative equity signals, which are rejected as cheap by
the rock-bottom calculations, all have quite low prices.
Figure 4 is an attempt to explain what is going on here in
a unified way. It charts the price of a bond against the
size of its equity cushion. For a particular bond, the
rock-bottom price is represented as a horizontal line,
because it is not affected by the size of the equity
cushion. Of course, a more accurate valuation of the
bond would depend on the equity cushion. We represent
this dependence by the put option price, so called
because it results from viewing the bond as a short put
option on the assets of the issuer, struck at the redemption value of the bond. From this point of view, a
decrease in equity value pushes the option further in the
money, and so translates into a decrease in the bonds
price. This bond price fall is larger, the smaller is the
initial value of the equity.
These two lines divide the chart into four regions in
which the market price of the bond can lie. In each one
we consider a bond whose equity price movement
(indicated by the arrow) makes it a Buy (A and B) or a
Sell (C and D), as opposed to a No View.
Our rule judges cheapness and dearness by comparing
market prices with rock-bottom prices, and will make
money if market prices gravitate to rock-bottom prices.
However, all other things equal, we would expect a
bonds market price to move toward the put option
price, and to move faster, the bigger the gap between the
two. Moreover, the lower the equity cushion, the more
the bond is like equity, and so the more we would expect
the bond price to move in tandem with the equity price.
The upshot of these presumptions is that the bonds we
encounter are likely to cluster more tightly around the
New York
October 25, 2001

Portfolio Research
equity line, the lower the equity value. The shaded area is
a fanciful designation of where the bulk of the bonds
would be found.
These conditions mean that type B and D bonds will tend
to be quite rare. Ds will evaporate because their sharp
drop in equity prices will translate speedily into a fall in
market price, which pushes them toward or into the
cheap region from the rock-bottom standpoint. Bs will
be rare because they should be quite close to the put
option line to start with, and so an above-average move
in equity price will tend to push them over to the other
side of the line. However, Ds are where the combined
equity-rock-bottom rule makes money on sells, while Bs
are where it loses money on Buys.

page 55
Figure 4
Bond prices and the equity cushion
Bond
price
UST price
D
B
C
A
Put option price
Rock
bottom
price
Equity cushion
The assumptions similarly mean that type A and C bonds

will be relatively stable. As are the bread-and-butter of
the Buy recommendations, while Cs are where the Sell
recommendations stand to lose money.
industry or maturity selection, or simply buying lowpriced bonds. Similarly, the rule performs well under
varying market conditions, both in High Yield and equity
markets, from which it draws its information.
It is the As and Ds that are the focus of the good news/

bad news contrast mentioned in the introduction. For
type A bonds, the equity cushion is large, and no remarkable positive events, such as record-beating earnings or
product innovations need to befall their equity for the
bonds to continue clipping coupons and outperform the
index. As long as their equity has outperformed, there is
the extra assurance that nothing negative has happened.
In other words, for As, no news is good news. The
situation of Ds is very different. They are already
expensive, and bad news that causes their equity price to
fall substantially lowers their equity cushion to a level
that causes the bond to trade more in line with equity,
and incorporate the same price movements. In other
words, for Ds, bad news travels fast.
The Sell recommendations of the basic rule do not

perform so well. However, a small tweak produces an
even simpler rule, just based on earlier equity returns,
whose performance since 1999 has been quite stellar. We
explain the rock-bottom signals negative contribution
here in terms of the relatively quick adjustment of the
High Yield market to bad news.
Taking stock
We have shown that a very simpleminded bondpicking
rule performed extremely well in the High Yield market
over the last two years. The rule removes bonds for
which information is inadequate, and assigns the remaining bonds to groups according to whether they score
above or below average on two quantitative measures an equity signal and rock-bottom valuation signal, both
based on fully public information. It recommends trading
on bonds whose signals tell a consistent story.
The Buy recommendations of this rule are very successful, and in no way resemble more familiar strategies of
So, the strategy that comes out of this would seem to be:
follow the Buy recommendations of the basic rule, i.e.,
buy bonds that the rock-bottom signal finds cheap, and
whose equity has outperformed; and sell anything whose
equity has fallen by more than 30% over the last six
months.
We are more sanguine about the first of these than the
second. Buying bonds that appear cheap and whose
equity does not make one suspicious is a way of exploiting quantitative anomalies in a market populated by bonds
with elaborate cashflow structures, and where quantitative relative value analysis is nevertheless not practiced
so widely as in, say, the US Treasury market. While
there is a large issuer-specific component to any bonds
price in the High Yield market, ultimately bond
prices reflect the size of promised cashflows, and the
chance of actually receiving them. By systematically
buying those bonds whose cashflows appear undervalued, the aim is to filter out the noise associated with
issuer-specific conditions.
New York
October 25, 2001

Portfolio Research
In contrast, selling bonds whose equity has fallen

sharply in the preceding six months, is harder to place
faith in, even though it has proved very successful. It is
the result of a small amount of datamining, as we
developed it after looking at the failure of the basic rules
Sell recommendations. But more important, it is hard to
understand how significant events that have been public
in the equity market for some time filter into bond prices
only later. Bonds and equity are both claims on the
underlying assets of an issuer. Equity may respond more
than bonds to news, but they should both respond at the
same time. So, in the absence of an economic reason for
why equity prices predict bond price movements, we
have less assurance that this connection will persist in
the future.
Postscript
After completing this analysis, we came upon reliable
equity price data that enabled us to examine the performance of the bondpicking rules from 1996II thru 1999I.
The results of buying bonds with above-average credit
ratios and prior equity returns are summarized in Table
7. Going back to 1998II, the story is much the same as
for the most recent two years: substantial
outperformance of the index. However, prior to that
time, the rules performance is erratic, and in no period
does the rock-bottom filter improve on the performance
of the equity filter.
Interestingly, the positive performance of the rule dates
to Russias devaluation and LTCMs troubles in mid1998. While many changes in spread markets occurred
at that time, the significant one for our rule is the jump in
the variability in performance across bonds. The smaller
is this variation around the index return, the less is there
differentiation in the fortunes of individual bonds, and
the harder it is for a bondpicking rule to outperform.
Prior to 1998II, the volatility of six-month returns over
the cross-section of bonds in the (vanilla) index averaged
6%. In 1998.II, it jumped to 13%, and has averaged
16% since. In the last two six-month periods, this crosssection volatility has been around 20%.
Loosely, one would expect an upward-trending market,
or even one driven by a bubble, to exhibit a low volatility.
This is consistent with the experience of the years 1996-

page 56
98I, when spreads narrowed substantially, and the equity

market seems to be a better guide to bond performance
than cashflow valuation. Only when the market turned
down, and performance of bonds became more varied,
did fundamental valuation of cashflows come into its
own. As we have seen, it held its own throughout the
indifferent market in High Yield of the last two years.
This is, of course, precisely the sort of time one would
want to be able to outperform.
Table 7
Annual returns: Buy recommendations of the Basic Rule

Vanilla Index
Buys
Equity Rule
Equity & RB
1996II 1997I 1997II 1998I 1998II 1999I

18.3
13.5
14.0
8.6
-2.6
4.8
19.4
18.3
12.7
12.7
15.8
15.0
10.8
9.8
6.7
8.5
7.2
8.6
Portfolio Research:
Peter Rappoport
Luca Brusadelli
Guy Coughlan
Alan Cubbon
Drausio Giacomelli
Halvor Hoddevik
Tatsushi Kishimoto
Canlin Li
Lee McGinty
Stephen Tang
Frank Zheng
(1-212) 834-7046
(44-20) 7325-5607
(44-20) 7777-1857
(44-20) 7325-5953
(1-212) 834-4685
(47-22) 941-978
(81-3) 5573-1521
(1-212) 834-9228
(44-20) 7325-5482
(1-212) 834-9224
(44-20) 7777-1534
(1-212) 834-9226
Related Research:
Valuing Credit Fundamentals: Rock Bottom Spreads,
November 17, 1999, P. Rappoport
Rock-bottom Spread Mechanics, July 11, 2001, P.
Rappoport
Introducing the Rock-bottom Roundup, July 11, 2001,
M. Sirinathsingh
New York
October 25, 2001

Portfolio Research
page 57
October 23, 2001
Portfolio Research
Picking Investment Grade Bonds

Valuation Rules
We describe a simple bond selection rule based on

each bonds rock-bottom price
its issuers recent equity performance
Figure 1
Outperformance of the Investment Grade Market by the Bond

Selection Rule
(Annualised Returns: six-month holding periods)
2.5%
The rule outperformed the investment grade market

by an average 1.1% annually since 1997
its information ratio was 1.2
1.5%
Although based on credit fundamentals, the rule is

successful because it mimics market technicals,
notably, the way investors express tactical views
Its simplicity and success underline the usefulness of
0.5%
quantitative credit valuation

equity market signals
-0.5%
1997II
Recommendations to buy or sell corporate bonds are

typically based on a view of individual companies, or a
view of an industrys prospects, or information about
supply pressures in maturity, quality or industry sectors.
They are the product of experience and qualitative
judgment, and require an intimate association with
market and economic conditions.
The bondpicking rule we describe here has a starkly
different origin. It takes a birds eye view of the market,
and requires no special knowledge. Its purely quantitative nature makes it possible to trawl through thousands
of bonds, valuing each in a common analytical framework. The data behind this valuation are just cashflow
patterns, credit ratings and bond and equity prices, all of
which are widely accessible. The only other input is
data-processing muscle.
We assess each bonds cheapness or dearness by
comparing its current market price with its rock-bottom
price1, which places a value on its promised cashflows,
given the issuers credit rating. This signal can be
misleading if the credit rating lags information that
1 This proprietary JP Morgan valuation framework provides a
price for any set of cashflows subject to credit risk. See the
references at the end of the paper for further information
1998I
1998II
1999I
1999II
2000I
2000II
2001I
Average
market prices already embody. So, we check the issuers recent equity returns, which should reflect any
significant change in its fortunes. Thus, the rule comes
down to the idea that, if the bonds rock-bottom valuation tells us it looks cheap based on its rating, and the
equity market provides no reason to suspect the rating,
then it probably is cheap. Similarly, it recommends
selling expensive-looking bonds whose equity has
underperformed.
Figure 1 displays the combined performance of the Buy
and Sell recommendations of this rule, over 6-month
holding periods. Since 1997, this strategy, in which
40% of the portfolio is turned over every six months
outperformed the universe of investment grade
corporates by an average of 1.1% annually. Its annual
information ratio over this period was 1.2. Sell recommendations over the same period returned an average of
1.0% annually, producing an information ratio of 1.3,
while the Buys annual average return was 2.2%, with an
information ratio of 1.1.
The next section of this research note describes how we
combine rock bottom valuations with equity data, to
arrive at the Buy and Sell recommendations. Then we
provide fuller details of their performance over the last
four years.
New York
October 25, 2001

Portfolio Research
The remainder of the note is devoted to understanding

why the recommendations have worked, and what they
tell us about the operation of the market. We document
that their performance cannot be dismissed as a run of
luck. Then we show that they are not a disguised
version of some more familiar strategy, such as selection
of industries or maturities.
The rules success derives from its ability to select
individual bonds to over- or underweight, rather than
sectors. This bond-specific, component indeed accounts for most of the differences in returns among
bonds. Industry effects are much smaller.
Many investors employ a bond selection strategy.
Typically, however, they analyse individual companies in
depth, taking into account many special and qualitative
factors. Our approach to bond selection is very different, relying on broad quantitative indicators, and differentiating among issuers only by reference to their credit
ratings. The question is, why does this form of bond
selection work?
Since our rule hinges on the gap between a bonds
market price and fundamental value (as measured by its
rock-bottom price), its success suggests that market
prices gravitate towards fundamental value. This kind of
effect can be significant very close to maturity. However
it is far too weak in the longer maturity bonds in which
the rule invests to account for its performance over
short timeframes.
Our preferred explanation is that while no individual
investor is necessarily carrying out this systematic
relative value analysis, their behavior in the aggregate tilts
toward favoring our Buys, and against favoring our
Sells. This comes about because investors who overor underweight sectors express their views by buying or
selling particular bonds, rather than all bonds in a
sector. In choosing these particular bonds, they will
tend to focus on value relative to fundamentals.
The performance of our rule lends credence to this
perspective, which we demonstrate by examining how
returns vary across bonds and industries. Indeed, the
amount of variation in returns attributable to bondspecific factors dwarfs that attributable to industryspecific factors and suggests that bond-selection, rather
than industry-selection is the activity most likely to pay
off.

page 58
All of this suggests that the simple quantitative indicators

we use rock-bottom prices and equity returns are
worth building into corporate bond strategy.
The bond selection rule
All other things equal, a bond is more attractive the
higher its promised cashflows, the lower its likelihood of
going into default, the greater its seniority, and the more
remote its chance of being called. The bonds rockbottom price explicitly accounts for these features, and
can be compared with its market price to assess whether
its credit exposure is cheap or dear. However, the rockbottom valuation describes the issuers credit only in
terms of its rating, which can lag the market. A bonds
market price may be below rock-bottom, not because it
is valued incorrectly, but because everybody knows the
issuer is to be downgraded, i.e., its real rating is lower
than the one used in the rock-bottom calculation. So it is
necessary to enforce the all-other-things-equal clause
strenuously.
Our rule uses changes in equity prices as a timely
indicator of issuers health. We build an index of the
equity returns of the issuers in our universe over the
preceding six months, which we use to assess each
bonds equity signal:
Positive equity signals: the issuers equity return over
the preceding six months exceeded the equity index
return;
Negative equity signals: the issuers equity return over
the preceding six months fell below the equity index
return;
With issuer-specific conditions summarized in this way,
we proceed to value the cashflows of each bond, by
calculating a credit ratio, defined as
credit ratio =
rock bottom price

market price
We use the credit ratio to partition bonds into two

groups, according to their rock-bottom signal
Cheap bonds: whose credit ratios exceed the average
(over all the bonds)
Expensive bonds: whose credit ratios fall below the
average
Buys: Cheap bonds with positive equity signals

Sells: Expensive bonds with negative equity signals
No View: bonds with conflicting signals (cheap
bonds with negative equity signals, and expensive
bonds with positive equity signals).
We used this rule to partition bonds into Buys and Sells
on the dates that marked the start of each of the eight
consecutive six month holding periods since mid-1997
(see Figure 2).
For the universe of bonds at the start of each period, we
took all investment grade corporate bonds tracked by
Bridge Data Systems whose issue size was no less than
$150 million, and for which prices were available both
on that date, and six months later. On average, this
universe comprised 2135 bonds. Of these, an average of
447 qualified as Buys each six months, and an average
of 543 qualified as Sells. For the remaining 1145 on
average, equity performance and credit ratios provided
conflicting signals.
Performance
We now present results on the rules performance over
the eight consecutive six month holding periods since
mid-1997. At the start of each period, we assume that
the investors portfolio is identical to the universe of
bonds described above, to which we refer as the
2 For the credit fundamentals forecast, we use historical default
and downgrade probabilities and historical recovery rates and
recovery volatilities. Different credit fundamentals forecasts
shift rock-bottom prices by approximately the same amount for
bonds in similar rating categories. Since we are interested in
relative value, the actual credit fundamentals forecast prevailing
at the time of selection is not critical, and thus we simply use the
historical values
Figure 2
Investment Strategy Timeline
Compare
market and
rock - bottom
prices
6 months of equity returns
Holding
period
1997.II
2000.I
2000.II
2001.I
6 month holding period
Trade on
recommendations
on...
July 1, 1997
Jan 2, 2000
July 1, 2000
Jan 2, 2001
Unwind
trades
on...
Dec 31, 1997
...
The rock-bottom price places a value on the pure credit

exposure of the bond, given its rating (and under a given
view of future credit fundamentals)2. These assessments
will only be reliable if the current rating is a correct
assessment of the issuers health. We have greater
assurance that apparently cheap bonds are cheap, if the
issuer also has a positive equity signal, because then
there is no reason to suspect that the rating will be
lowered. Similarly, apparently expensive bonds of issuers
with negative equity signals are more likely to be expensive, because there is less likelihood of an imminent
upgrade. This reasoning leads us to the final recommendations of the basic rule:

page 59
...

Portfolio Research
...
New York
October 25, 2001
June 30, 2000

Dec 31, 2000
June 30, 2001
index. Recommended bonds are then bought and sold.

The volume of purchases each period is set at 40% of
the value of the portfolio (80% annually), which is about
average for investment grade fund managers. The
Buys are financed first from liquidation of recommended Sells, and then, if necessary, from the remaining bonds in the portfolio, in proportion to their market
capitalization. At the end of the six-month holding period,
the investor unwinds all transactions, returning to the
neutral index portfolio. Figure 2 shows this timeline for
the most recent four holding periods. To calculate
returns, we use bid prices, subtracting a standard bidoffer spread ($0.46 per $100 face) for each of the two
round-trips of transactions involved.
In order to isolate the pure credit component of the
rules performance, we hedge out interest rate risk, as
represented by the swap market. That is, from each
bonds six-month total return, the return on (receiving
fixed on) a swap of equal duration has been subtracted,
leaving only pure credit returns. Thus, for example, the
index return we report is the market-capitalization
weighted aggregate of the credit returns of all of its
constituent bonds.
Table 1 provides information on the returns of the entire
index and baseline strategy. The baseline portfolio
strategy of financing a 40% investment in the Buy
recommendations from Sells and No View bonds,
outperformed this vanilla index by 1.12% annually, on
average.
Table 1 also decomposes the investment results into
Buys and Sells. The Buy recommendations only
Accounting for the rules success

We now document that our Buy recommendations do
not boil down to some other, more familiar strategy. One
possibility is that we are carrying out sector selection
under another name. That is, does our rule outperform
because it picks bonds whose industries outperform? To
Tracking
Error
Information
Ratio
1.3
7.3
6.7
-0.1
2.1
1.12
1.0
1.16
Equity
Equity & RB
-0.3
-0.2
0.9
4.3
0.49
2.17
0.5
1.9
1.02
1.12
RB Only
-0.9
2.8
0.89
1.6
0.55
Equity
Equity & RB
RB Only
-0.2
-0.4
-0.9
0.8
1.3
1.9
0.44
0.99
0.86
0.4
0.8
1.2
1.07
1.28
0.72
Index return
Outperformance
Baseline
Strategy
Buys
Sells
Minimum and maximum returns are on a 6-month basis, while the

other statistics are annualised. Combined strategy figures account
for transactions costs, while those for Buys and Sells do not.
Table 2
Success rates (% of total number of bonds)
Outperformers
% of Index Bonds
relative to index
% of Buys
Equity
Equity & RB
RB Only
Underperformers
% of Index Bonds
relative to index
% of Sells
Equity
Equity & RB
RB Only
Average
Tables 1 and 2 also illustrate the interaction between the

equity and rock-bottom parts of the rule. Just picking
bonds on the strength of recent equity performance
would not have produced particularly stellar results,
especially in terms of successful bond picks, rather
than returns. The equity only rule averaged a 3%
higher success rate than the Index for both Buys and
Sells. However, this does not mean that it is dispensible,
because the RB only rule also underperformed the
combined Equity and RB rule. This is consistent with
our idea that the use of equity information is to provide a
double-check that an issuers rating is an up-to-date
reflection of its credit quality.
Average
These are reasonable success rates for active management, but they are highly statistically significant, given
the number of bonds involved. There is less than a 1-in50-billion chance that the success of the Buy recommendations (relative to the 58% success rate of the
index) is due to chance. The corresponding probability
that the Sell recommendations are due to chance is 1 in
10 million. These estimates are very conservative, as
they are based on the average figures for a single period.
There is a much smaller probability that the actual
success rates of our Buy and Sell recommendations
could be repeated on average for eight periods in a row,
were they merely picking bonds at random from the
index.
(% returns)
Best 6
months
Since it selects bonds irrespective of their size, more

precise information on how the rule works, and where
its strengths and weaknesses lie, comes from looking at
its performance on a headcount basis. Table 2 shows
that, in an average six month period, 58% of bonds
outperformed the index return. Our Buy picks based on
the combined equity and rock-bottom rule did 15%
better. Similarly, our combined Sell rule picked 11%
more underperformers than the indexs 42%.
Table 1
Credit returns and outperformance
Maximum
underperform the index once, and return significantly

more than the Sells. However, the Sell recommendations produce an annual information ratio of 1.28
(before transactions costs).

page 60
Worst 6
months

Portfolio Research
Minimum
New York
October 25, 2001
38
77
58
-7
-10
-16
8
44
36
3
15
9
23
62
42
-7
-8
-12
8
24
22
3
11
8
answer this question, we decompose the overall return

outperformance of the Buy recommendations into two
components:
Industry selection: how much we would have outperformed the index had we built a portfolio with the
industry breakdown of our Buy picks, but with the
same bonds within each industry as the index.
Bond selection: Using the Buy rules weighting of each
industry, this aggregates the Buy rules outperformance
Insurance
Leisure
Manufacturing
Land dev./Real Estate
1.8
0.6
Mobile/Modular Homes
0.4
0.1
Beverages
9.1
9.6
Cosmetics/Toiletries
2.8
3.5
Food
19.0
20.3
Home Furnishings
0.6
0.3
Household Products
1.6
0.3
Leather/shoes
1.0
0.2
Retail Stores
23.9
25.2
Oil and Gas
41.9
31.3
Banking
98.1
88.2
Finance
101.8
126.8
Investment
8.8
5.3
Securities
14.5
39.3
Drugs-Generic and OTC
10.6
8.0
Health Care Centers
2.0
1.1
Medical equipment/supply
6.3
2.9
Data Processing
6.8
9.5
Electronics/Electric
17.8
13.4
Insurance
26.8
13.4
Filmed Entertainment
1.1
1.3
Leisure/Amusement
12.1
9.0
Auto parts/equipment
4.8
2.4
Auto/Truck mfrs.
4.8
13.6
Chemicals
17.6
9.1
Coatings, paint, varnishes
1.6
0.4
Containers
2.9
1.2
Glass/products
2.1
1.2
Machinery
14.0
7.5
Manufacturing/Distr
4.5
3.6
Office Equipment
1.5
1.0
Paper/Products
14.9
10.2
Plastic/Products
7.0
6.5
Rubber
3.0
1.1
Specialty instruments
0.8
1.0
Textiles
1.9
0.4
Tobacco
1.9
1.2
Media/Telecom Advertising/Communications
0.1
0.1
7.4
8.3
Broadcasting
Metals/Mining
Other
Service
Transport
Utility
Graphic Arts
1.8
0.5
Publishing
13.5
16.0
Telecommunications
44.0
51.4
Aluminum
3.3
1.2
Metal
0.5
0.2
Mining/Diversified
9.4
3.0
Steel-Iron
2.4
1.0
Conglomerate/diversified
2.8
2.4
Foreign
1.0
0.6
Pollution Control
1.4
1.2
Real Estate Investment Trust
21.9
7.1
Food serving
2.0
0.6
Hotels/Motels/Inns
1.4
0.8
Services
3.6
1.5
Air Transport
6.0
5.5
Auto rental/service
1.8
2.2
Railroads
9.1
13.9
Transportation
4.4
2.2
Trucking
0.5
0.1
Util.-Diversified
11.6
7.7
Utilities-Electric
36.4
25.5
Utilities-Gas
24.5
15.4
The industry designation of each issuer is as determined by Standard

and Poor
8%
Bond Selection
6%
Industry
Selection
4%
2%
0%
-2%
2001I
2.6
Average
HiTech
12.9
7.8
2000II
Health
8.1
Building
(% per 6 months)
2000I
Energy
Finance
Aerospace
Industry performance attribution of buys
1999II
Consumer
Average Capitalization
($billions)
1999I
Aerospace
Building
Average
Issuers
Figure 3
1998II
Table 3
Industry composition of investment grade bonds 1997-2001

page 61
1998I

Portfolio Research
1997II
New York
October 25, 2001
of the index in each industry. So, if the Buy rule

selects no bonds in a particular industry, that industry
contributes zero to bond selection.
Table 3 shows how the issuers in our bond universe are
distributed among industries. Figure 3 shows the contributions of bond and industry selection in each holding
period. Only 0.4 percentage points of the 2.2% average
outperformance of the Buy recommendations comes
from industry selection, the rest coming from bond
selection. In absolute value terms, the bond selection
component typically exceeds the industry one by a large
margin. Thus, our Buy recommendations are not
mimicking an industry selection strategy.
A similar analysis shows that our Buy recommendations
do not reduce to advocating buying certain maturities.
The maturity profile of the portfolio of Buys is very
close to that of the high grade universe. Likewise, neither
large or small issue sizes are represented among the Buys
differently from the index as a whole.
The main drivers of rock-bottom prices are the rating
and maturity of bonds. The only other sources of
difference among bonds rock-bottom prices are coupons and seniority. So, once the rating and maturity
have been controlled for, there should only be minor
differences in rock-bottom prices between bonds.
Maybe the rock-bottom filters designation of cheap and
dear bonds amounts to nothing more than recommending
the purchase of bonds that are cheap within each rating/
New York
October 25, 2001

Portfolio Research
maturity bucket (for example, Baa1s with 5-7 years until

maturity ).
To examine this possibility, we look at the performance
of bonds whose equity outperformed in the preceding
six months, and whose market prices are below the
average of their rating/maturity subsector. So there is no
rock-bottom calculation involved here at all.
A low-priced bond is defined as one whose market price
is below the average of the rating/maturity bucket it
occupies. Rating/maturity buckets were constructed
from the ten ratings: Aaa, Aa1, Aa2, Aa3, A1, A2, A3,
Baa1, Baa2, Baa3, and the five maturity buckets: 1-3yrs,
3-5, 5-7, 7-10, 10+.
A Buy rule that overweights the cheap bonds in each of
these fifty categories (10 rating categories * 5 maturity
buckets) underperforms the index by 0.01% every six
months, which is again dwarfed by the 1.1%
outperformance of the rock-bottom/equity rule. This
shows that the rock-bottom rule outperforms not only
by selecting cheap bonds within the rating maturity
buckets, but also by incorporating a valuation between
these buckets. Just buying cheap bonds has nothing to
offer in this respect.
Our Buy recommendations thus seem to identify a
source of value in the investment grade market that does
not boil down to any of the other obvious suspects.
The question of data quality
One possible explanation for the success of the bond
picking rule is that the underlying bond price data
contain errors. For example, a bonds price may have
been entered (by a contributor to the historical database)
at the start of one of our investment periods as $59.21
instead of $89.21. This would cause the bond to
exhibit a very high credit ratio, which need not be
contradicted by its prior equity returns. Consequently, it
becomes a Buy recommendation. During the holding
period, the bonds price is corrected, to bring it into the
range of $89.21. Our accounting scores this as a $30
appreciation, and a big success, whereas no such
opportunity existed in the actual market.
While the low reliability of price data is notorious in the
High Grade market, we find no evidence to suggest that
it accounts for the success of our bond selection rule.
We discuss two types of information that lead us to this
conclusion.

page 62
Figure 4
Outperformance of Buy recommendations at 3-, 6-, and 12month horizons (% returns, not annualized)
5%
4%
1-yr holding period
6-month
holding period
3-month
holding
period
3%
2%
1%
0%
1997II
1998I
1998II
1999I
1999II
2000I
2000II
2001I
Average
to 2000II
-1%
First, to be significant, these pricing errors need to

produce extreme returns (positive for Buys, negative for
Sells). So, examining the performance of Buys and Sells
after removing the extreme performers should significantly affect average returns. But removing both the top
and bottom performing 5% of bonds from each brings
about very little change in average returns. For Buys,
the biggest change is a decline of 0.19% in 2001I. And
the average change (over all eight periods) is actually
positive for Buys, and negative for Sells, indicating that,
in aggregate, removing the tails has the opposite effect
from that expected if pricing errors drove the results.
Second, for one period, 2000II, we examined the
monthly price movements of each bond to see if a
sudden jump occurred for any of our Buys and Sells.
Not one bond evidenced this pattern. Thus, while we do
not believe that all of the prices we use are a perfect
reflection of trading opportunities at the time, we have
no reason to believe that data problems account for the
success of our bond selection rule.
Could investors be using our rule consciously?
Our rule selects bonds whose market prices are at odds
with their fundamentals valuation. Because it outperforms the universe of bonds, there must be some
tendency for the prices of the selected bonds to adjust
towards their fundamental values. Figure 4 shows that
this outperformance occurs not only after 6 months, but
also at a 3-month and 1-year time horizon.
While Figure 4 documents a substantial pull towards
fundamentals, it is hard to believe that this occurs
New York
October 25, 2001

Portfolio Research
because investors are systematically comparing credit

fundamentals with market prices, and acting in line with
our rule. There are two reasons.
First, we found that the rule does not replicate other,
more familiar approaches to picking bonds. The indicators it uses certainly do not figure prominently in the oral
tradition of tactical management. While we found that
most of its outperformance comes from bond selection,
this is not the kind of qualitative, intensive analysis of
company specifics that most associate with the term.
Second, the adjustments indicated in Figure 4 are very
large. To see this, we need to return to the credit ratios
we have calculated for bonds, defined on page 3, and
repeated here, for convenience:
For a bond to be a Buy at any point in time, it is necessary (but not sufficient) that its credit ratio exceed the
credit ratio =
rock bottom price

market price
average credit ratio over all bonds at that time. Any Buy
that subsequently outperforms the index will see its
credit ratio pushed down toward the average. (Most of
the variation over time in credit ratios comes from
market price changes, since rock-bottom prices change
little over , say, one year). And, once the bonds credit
ratio falls below the average, it is no longer a Buy. How
much do Buys market prices have to rise for them no
longer to qualify as Buys in this way? The average is
1.7%, over all Buys in all investment periods for which
we have a years experience (i.e., all except 2001I).
Figure 4 shows that more than half this gap (1.1%) is
closed by the average Buy over the space of one year .
This seems excessive.
Does the market follow the rule unconsciously?

If the market is not consciously following the rockbottom framework, could it be doing so unconsciously?
A common strategy among corporate bond investors is
the top-down approach illustrated in Figure 5. This
involves first picking industry sectors to over- or
underweight. They then express these views by
transacting in individual bonds. If they want to over-

page 63
Figure 5
The top-down approach to corporate bond selection
Market
Neutral
Industries
Overweight
Industries
Industry
Selection
Underweight
Industries
Buy best
bonds in
industry
Bond
Selection
Sell worst
bonds in
industry
weight a sector, they buy bonds that look like good value
in the sector, which means that the bonds appear cheap,
given their rating, maturity, and coupon. Underweights
are accomplished similarly by selling dear-looking bonds.
Figure 5 describes the behavior of an individual investor.
The pattern of returns we observe in the market will be
the aggregate of this kind of behavior across all investors.
The performance of each industry will depend on the
relative numbers of (amounts demanded and supplied by)
investors seeking to under- or overweight it. While all
investors may have strong views about an industrys
prospects, the overweights and underweights can quite
plausibly offset each other, with the result that industry
average returns differ little from index returns.
The performance of each bond depends on whether there
is an excess of investors wishing to sell it or to buy it. In
contrast to the case of an industry, these effects will tend
to go in one direction, rather than cancel out. This is
because investors looking to underweight an industry will
try to identify expensive-looking bonds, while those
seeking to overweight the industry will seek cheap-looking
bonds. What makes a bond look cheap to one investor
for example, good equity performance, combined with a
low price relative to bonds with the same rating and
maturity, is unlikely to make it look expensive to another
investor.
New York
October 25, 2001

Portfolio Research
Figure 6
Effects of industry views on performance when investors

select bonds according to the top-down approach
Industry
View
Performance vs Index
BestIndustry
looking
bonds
Overweight
Consensus
Underweight
Consensus
Varied (50/50)
Worstlooking
bonds

page 64
Figure 7
Sources of market variation of returns

3.5%
2.5%
1.5%
Market variation
Bond variation
0.5%
Industry variation
0
1997II
From this there emerges very little to guide one in terms

of selecting industries it is matter of second-guessing
the popularity contest among investors. In contrast, the
guidance for picking bonds is very clear. Any bond with
good credit fundamentals will have a good chance of
being bought by some investor looking to overweight its
industry. Any bond with bad credit fundamentals will
have a good chance of being sold by some investor
looking to underweight its industry. So, it makes sense
to buy the bonds with good credit fundamentals, and to
sell the ones with bad credit fundamentals. And, unless
you have expertise in predicting which industries will be
predominantly over- or underweighted, this should be
done irrespective of the bonds industry.
We would expect this bondpicking strategy to be more
successful, the less consensus there is in industry views.
If all investors are bullish about Aerospace, then no
expensive Aerospace bonds will be sold, and our rules
Sell recommendations that happen to be in Aerospace
will not do well. Similarly, a bearish consensus will
result in no buying of cheap Aerospace bonds, causing
Buys to underperform. These outcomes are illustrated
schematically in Figure 6.
While we cannot observe investors industry views, we
can observe the patterns of returns that are the result of
their views in aggregate. These patterns suggest that the
Varied scenario is more the rule for aggregate industry
views than Consensus. Over our eight investment
periods, bonds that moved counter to their industries
performance were quite common. In industries that
underperformed the index, 41% of bonds outperformed
1998I
1998II
1999I
1999II
2000I
2000II
2001I
the index. In industries that outperformed the index,

31% of bonds underperformed. In 95% of industries,
more than 10% of the bonds moved opposite to their
industries (for example, they outperformed when their
industries underperformed). Similarly, in 81% of industries, more than 20% of bonds moved contrary to their
industries. In each case, the top-down rule leads us to
view the underperformers as likely Sells of investors
wishing to underweight the bonds industry, and to view
the outperformers as the Buy picks of investors who are
bullish about the bonds industries. So there was
scarcely consensus.
Another implication of absence of consensus is that
industry average returns will be close to index returns, as
a result of bullish and bearish investors in each industry
cancelling out each others effects. So, the range of
variation of bonds returns should be large compared to
the range of variation of industries returns.
To develop the relevant evidence, we need to separate
the performance of individual bonds from the performance of the industry sectors in which they reside. We
define market variation as the standard deviation of
returns across all bonds in the market in each period.
We break this market variation into two components.
Industry variation, which is the standard deviation of
the average returns in each industry. In other words,
to calculate industry variation, calculate the average
return for the bonds in each industry, then calculate the
standard deviation across these industry averages.
Bond variation, which measures the range of indi-
New York
October 25, 2001

Portfolio Research

page 65
Figure 9
Figure 8
Outperformance: Rock-bottom rule Buys and Sells
Outperformance:
pure industry selection with perfect foresight
Market
Rock-Bottom rule
0.7%
Bond vs index
Industry type
UnderOutRates of
Occurrence performer performer
Outperforms
70%
41%
Underperforms
30%
59%
Success
rates
Bond position
1.5%
Industry type
OutUnderperformer performer
Buys
79%
54%
Sells
40%
64%
Average
2001I
2000II
2000I
1999II
1999I
1998II
1998I
-0.5%
1997II
0.5%
vidual bond returns around their respective industry

averages. To calculate bond variation, calculate the
deviation of each bonds return around its respective
industry average, then calculate the standard deviation
of these deviations.
The relationship among these three measures of return
variation across the population of bonds is:
Figure 7 tracks these measures for the eight six-month
2
Market
Industry
Bond
=
+
Variation
Variation
Variation
periods we have been examining. While all three are

positively correlated, most of the changes in market
variation are reflected in bond variation, as the predominance of varied industry views would lead us to suspect.
Since industry variation makes up a small part of total
market variation, one would not expect to make a great
deal from a pure industry strategy, that is, one that
over- or underweighted all bonds in an industry in
proportion to their market weights (so that no bond
selection is made). Figure 8 provides an upper limit on
the return of this kind of strategy. It shows how much
one would have earned in each six-month investment
period, if one had perfect foresight about industries. To
be precise, in each period the strategy overweights all
industries that ended up outperforming the index, for a

total of 40% of portfolio value. These overweights are
funded by selling industries that underperformed, and
the standard $0.46 round-trip transactions cost is used,
so this rule is structured the same as the Buy/Sell
version of our rule in Figure 1. The perfect foresight
pure industry rule averages 0.7% annually, compared
with our rules 1.1%.
Of course, the top-down approach to bondpicking is
not a pure industry rule. But this result suggests that
investors who use it are more likely to earn from the
bond selection step than from the industry selection
step. Indeed, as mentioned above, this is what our
rock-bottom rule does. It effectively skips the industry
selection step, and just picks bonds with extreme credit
fundamentals, irrespective of their industries. Figure 9
shows that bonds selected using this approach also
outperform, irrespective of the fortunes of their
respective industries. The left panel divides the bonds
in the market, first according to whether their industries out- or underperformed the index, and second,
within each industry, into out- and underperformers.
So, for example, in the average outperforming industry,
70% of the bonds outperformed, and 30%
underperformed.
The purpose of our rule is to try to identify these
outperformers as Buys and underperformers as Sells.
As the right panel shows, the rule accomplishes this
irrespective of the performance of the bonds industry,
and irrespective of the performance of the bond. For
example, it can pick outperformers in outperforming
industries with a 79% success rate, which is 9 percentage points better than the 70% rate one would expect
(left panel) if one had no ability. It can pick
outperformers in underperforming industries thirteen
percentage points better than pure luck (54% - 41%).
New York
October 25, 2001

Portfolio Research
Similarly, it does better than the no-ability success

rates in picking underperformers.
In summary, our rule is a pure bond selection rule. At
first glance, the fact that it outperforms presents a
puzzle, because investors do not explicitly look at rockbottom valuations across all bonds, yet they must be
moving bond prices in a way that is consonant with the
bond selection rule. Moreover, each investor tends to
follow a top-down rule that emphasizes industry
selection as well as bond selection.
This puzzle is resolved by looking at the effects of
investors as a group following the top-down rule.
Industry effects will tend to be muted in aggregate,
because overweight and underweight investors in the
same industry will tend to cancel out. Indeed a minimal
amount of actual market variation in returns is traceable
to differences in performance among industries. What
survives the aggregation is individual investors bond
selections. Bonds with good fundamentals will tend to
be in demand, as a result of the activities of some
investors seeking to overweight the bonds industries,
and bonds with bad credit fundamentals will tend to be
in excess supply, because some investor is selling them
to express a negative industry view. So a rule that buys
bonds with good credit fundamentals, and sells those
with bad credit fundamentals, irrespective of industry,
should outperform, irrespective of industry, as it does.
Related Research:
Valuing Credit Fundamentals: Rock Bottom Spreads,
November 17, 1999, P. Rappoport
Rock-bottom Spread Mechanics, August 1, 2001, P.

page 66
Conclusion
A popular distinction in market strategy is between

strategies that are based on fundamentals, and those
that are based on technical conditions in the market.
Our rule is overtly based on fundamentals. Its performance would seem to advocate a fundamentals strategy: if
a bond looks good according to the rock-bottom valuation, and the equity market provides no reason to smell a
rat, then the bond probably is a good deal.
While following this rule over the last few years would
have turned in a handsome outperformance, we find it
improbable that the market was consciously responding
so promptly to fundamentals information. Instead, we
trace the rules success to more technical considerations,
namely, the way investors implement their sector views.
As long as investors continue to express sector views by
buying or selling particular bonds, and as long as our
rules measures of fundamentals remain appropriate, the
rules success can be expected to persist.
Moreover, our discussion of the technical reasons behind
the rules success suggest that it is may have limited
downside: while it may not always make money, it will
not underperform by a large amount. The worst case
will occur when investors views exhibit a consensus on
which industries to over- and underweight. In these
periods, our rule will tend to make money from Buys in
industries that outperform the index and from Sells in
underperforming industries. It will lose money from
Sells (Buys) in the outperforming (underperforming)
industries. But there is no presumption that the winning
recommendations will outweigh the losing ones, or viceversa. The best guess is that they will be a wash. So,
in periods with high consensus, the expectation is that
the strategy will average a zero return, or maybe a little
less, due to transaction costs.
New York
October 25, 2001

Portfolio Research

page 67
May 4, 2001
Fixed Income Strategy
Valuing rating-triggered step-up bonds

Rating-dependent coupons add to the value of
bonds such as the most recent DT and BT issues,
but to a degree that critically depends on:
-
the exact form of rating dependence
bond maturity
current rating
These results flow from the JPMorgan Rock

Bottom Spread framework
During the past year, telecom operators such as BT, DT,
and KPN issued bonds that included step-up language to
protect investors in the event of a downgrade. The
bonds pay coupons which depend upon the issuers
current rating, as shown in Chart 1. Coupon step-ups are
in effect credit options, but valuing these options using
traditional derivative methods is difficult. The main
questions addressed in this piece are: (1) how should
bonds with ratings-triggered coupon step-ups be valued
and (2) how is the spread on these bonds likely to
change in the event of downgrades.
presence or absence of a step-up clause, we can value the

step-up by comparing the credit spreads on two identical
bonds, one with a step-up clause, and one without.
Our previous work on pricing the fair compensation for
bearing credit risk (Valuing Credit Fundamentals: Rock
Bottom Spreads, November 1999) provides a natural
framework for measuring the credit component of a
bonds market spread, or what we term its Rock Bottom
Spread (RBS). Rock-bottom spreads are essentially
reservation spreads for credit risky bonds. This
valuation draws only on credit fundamentals data that are
essentially external to the market, such as default rates
and recovery rates, and on the investors risk tolerance.
Basically, the RBS framework values a bond by tracing
its future possible credit ratings and combining the
probability of these ratings occurring with what the bond
will pay given each of these future ratings. A traditional
bonds cashflows do not vary from one rating to another,
except when the bond defaults. Capturing the effect of
cashflows that change in a prescribed way as the rating
changes is a simple amendment to this framework.
Chart 1
Applying the RBS methodology to the BT and DT bonds

gives the rock-bottom spreads for these bonds.
Removing the step-up provision and applying the RBS
analysis gives the rock-bottom spreads for plain bonds.
The difference between these two numbers is the value
of the step up. These values are presented in Table 1.
Coupons step up when ratings decrease

(%)
13
12
BT 7.625%
11
Table 1
10
Value of step-up clauses for BT and DT bonds
9
8
DT 7.75% $05
7
6
Aaa
Aa3
A3
Baa3
Ba3
B3
Market spreads can generally be regarded as

compensation for two different risks credit risk and
liquidity risk. Thus, we can think of partitioning the
market spread into two discrete components, a credit
spread, which compensates for the probability of loss
due to defaults, and a liquidity spread. Assuming that
the liquidity risk of a bond is not affected by the
BT
BT
BT
DT
DT
DT
Bond
7.625%
8.125%
8.625%
7.750%
8.000%
8.250%
$05
$10
$30
$05
$10
$30
Rating
A2/A
A2/A
A2/A
A2/AA2/AA2/A-
Value of step-up
16 bps
27 bps
44 bps
7 bps
10 bps
14 bps
These results assume a general credit fundamentals view

of a 34% recovery rate on defaulted debt, and an
investment grade downgrade rate of 12% for the next
New York
October 25, 2001

Portfolio Research
year, followed by the historical rate of 7% for the

remaining years. The BT bonds step up 25bps per rating
notch per agency below A3/A-, and the DT bonds step
up 50 bps if ratings go below A3 and A-, as shown in
Chart 1. Coupons step down in the exact opposite
manner if the bonds are upgraded.
This explicit valuation of the step-up clause makes it
easier to determine what the market spreads for these
bonds should be. For example, if a plain 10yr bond, with
an 8.125% coupon and credit quality similar to that of
BT is trading in the market at T+280 bps, we would
expect the 10yr, 8.125% BT bond to be trading
somewhere near T+253 bps, as a result of the step-up
clause.
S&Ps and Moodys negative outlook on DT prompts
the question of what the step-up clause would be worth
if the company were downgraded. To evaluate this, we
calculate the RBS of the bonds as if they had a lower
rating, and subtract that from the RBS of a bond without
the step-up. In each case we compare the credit spread
on the step-up bond, with a plain bond having the same
coupon as the step-up bond would if it had been
downgraded. For example, to value the step-up clause
for the DT 8% $10, if it were downgraded to Baa1, we
compare it to a plain bond having a coupon of 8.5%, as
this is the coupon the DT bond would pay if it were rated
Baa1. Table 2 gives the results of this analysis.
Table 2
Value of step-up clauses for downgraded bonds (bps)
BT
BT
BT
DT
DT
DT
Bond
7.625%
8.125%
8.625%
7.750%
8.000%
8.250%
$05
$10
$30
$05
$10
$30
A2
16
27
44
7
10
14
A3
33
47
60
15
18
20
Baa1
22
32
41
-10
-10
-12
Baa2
13
18
23
-5
-7
-9
As is evident from Table 2, the step-up clause has the

greatest value in the A3-rating state. This is not
surprising as it is in this state that there is the greatest
expected benefit from the step-up clause. Somewhat
surprising at first glance is the negative value for the DT
bonds in the Baa1 and Baa2 states. This is easily
understood when one considers that relative to a plain

page 68
Baa1-rated bond, the DT bonds have a step-down

clause, i.e. if the bonds are upgraded the coupon
decreases, while further rating downgrades have no
effect (see Chart 1).
Using the RBS framework and the above results we may
also evaluate how the spreads on the step-up bonds are
likely to change if the bonds are downgraded. Consider
a plain A2-rated, 10yr, 8.0% bond, trading at a spread of
280 bps. Assuming liquidity spreads remain constant
and given the general credit fundamentals view
mentioned above, the RBS framework suggests that this
bond should trade at 294 bps, if downgraded to A3, for a
spread widening of 14 bps.
Using Table 2, the value of the step-up clause for the DT
8.0% $10 is 10 bps, if DT is rated A2, and 18 bps if DT
is rated A3. Hence this bond should trade at a spread of
270 bps (280 bps 10 bps), if rated A2 and T+276 bps
(294 bps 18 bps), if rated A3. This corresponds to a
spread widening of 6 bps. Thus on the plain bond, the
spread widens by 14 bps on downgrade, while the DT
bond widens by 6 bps. This analysis for the other bonds
is shown in Table 3.
Table 3
Spread widening due to downgrades (bps)

Downgrade from: A2 to A3 A3 to Baa1 Baa1to Baa2
Plain 7.750% $05
8
16
19
BT 7.625% $05
0
28
28
DT 7.750% $05
0
41
14
Plain 8.000% $10
14
19
25
BT 8.125% $10
0
34
38
DT 8.000% $10
6
48
20
Plain 8.250% $30
18
22
26
BT 8.625% $30
2
40
44
DT 8.250% $30
13
53
21
Thus, while the step-up clause affords some protection

against spread widening given an A2 rating, it actually
causes the spread widening in successive downgrades to
be more severe than that of a plain bond.
To summarize, our analysis indicates that the step-up
clause adds value given an A2 or A3 rating for both BT
and DT bonds, while it actually makes the DT bonds
cheaper given a Baa1 or Baa2 rating . Furthermore, the
step-up clause affects the way the market spread on these
bonds changes in downgrades thereby making them
more or less attractive than plain bonds depending upon
their current rating.
New York
October 25, 2001

Portfolio Research
David Xu

page 69
February 2, 2000
Portfolio Research
Comparing Credit Fundamentals:

Emerging Markets versus High Yield
Emerging-market and high-yield spreads currently fall
short of spreads warranted by our credit fundamentals
views; the shortfall is much larger in high yield
Both markets now trade close to the spreads implied by an
optimistic view of credit fundamentals
Strategic investors should focus on long-term creditfundamentals volatility, rather than short-term market
volatility. This long-term perspective casts emerging
markets risk-return tradeoff in a more favorable light.
Introduction
Emerging market sovereign paper and speculative-grade
U.S. corporate bonds are both used by investors to
achieve significant credit exposure. This is practically the
only feature the two asset classes have in common.
Sovereigns and corporates involve very different packages
of credit fundamentals: exposure to changes in credit
quality, default, and subsequent recovery. These
differences make it very difficult to compare their value
and, consequently, to make high-credit risk investment
decisions. For example, spreads on comparable high-yield
corporate and emerging market sovereign indices are 451
bps and 587 bps, respectively. Is this 136 bps difference
sufficient to compensate for the lower diversity and lower
recovery rates of sovereigns among other differences? Is
it too much, given sovereigns higher average credit
quality? Without explicitly valuing credit fundamentals, we
simply cannot say.
Our analysis places high yield and emerging markets on
the same footing, by translating their credit fundamentals
into a spread investors should demand for each exposure.
At these rock-bottom spread levels, the asset classes
deliver a target excess return over Treasurys per unit of
risk, or information ratio. This framework has been laid
out in a recent research report1*, and a rock-bottom spread
calculator is available on our Morgan Markets website.
We compare high yield and emerging market indices that
contain securities of roughly similar structure:
*
Numbers refer to notes at the end of the paper.
Merrill Lynchs High Yield Cash Pay index (MLHY, for

short)
J.P. Morgans EMBI+, excluding defaulted issues and
collateralized issues such as some Brady bonds

( EMBI px+ for short; p stands for performing, x
stands for excluding collateral)
Table 1 displays rock-bottom and market spreads for the

two indices. EMBI px+ credit fundamentals require some 71
bps more in spread to deliver the same information ratio.
The EMBI px+ market spread exceeds MLHYs by 136 bps.
Thus, while both indices currently pay less than rockbottom, the EMBI px+ shortfall is much smaller.
Table 1
Market spreads and Rock-bottom spreads

Basis points (as of close of January 26, 2000)
Market
Rock
bottom
Difference
+
EMBI
px
EMBIpx
587
604
-17
MLHY
451
533
-82
Difference
136
71
To arrive at these rock-bottom spreads, we need to

account explicitly for differences in the indices credit
fundamentals:
Credit quality: The MLHY has slightly more lower
quality paper, and lower maturity
Diversification: The MLHY is spread more broadly
across issuers
Recovery rates in default: Less is expected from
sovereigns than corporates
Credit migration: Downgrades and defaults are less
frequent for sovereigns than corporates
Anticipated credit trends: Defaults are expected to
remain above historical averages for high yield
corporates, while emerging market sovereign upgrades
are expected to continue.
New York
October 25, 2001

Portfolio Research
Evidently, the second and third of these differences favor

high yield, while the others favor emerging markets.
Their combination nets out to the 71 bps difference in
rock-bottom spreads. These figures are based on our
best-guess scenarios for credit fundamentals. They are
influenced both by historical experience and our current
view of credit trends.
It is also possible to estimate upper and lower limits for

warranted spreads from pessimistic and optimistic
views of credit fundamentals, respectively. These are
illustrated in Figures 1 and 2. Market spreads of both
indices are currently close to their optimistic scenario
lower limits.
Figure 1
EMBI p+x spread and optimistic and pessimistic scenario levels

Basis points

page 70
have faced in the past (although the past six months have
been very tranquil). This market volatility greatly
exceeds that emanating from credit fundamentals, which
means that the risk-return tradeoff improves dramatically
for emerging markets as the investors time horizon
lengthens. At short time horizons, the tradeoff appears
more favorable to high yield. However, whether or not
the necessary liquidity exists to move portfolios in and
out of the high-yield market remains an issue.
In the next section of this report, we lay out the
ingredients of emerging market and high yield credit
fundamentals that go into our baseline rock-bottom
spread estimates. Then, we describe how we arrive at
the spreads that inform our optimistic and pessimistic
scenarios. Finally, we show how the risk-return
tradeoffs of the two asset classes are materially affected
by the investment time horizon.
1600
1400
1200
1103
Quantitative drivers for MLHY and EMBI p+x have

different characteristics
1000
800
600
587
453
400
In emerging markets and high yield markets, we identify

the following five distinctions as the key drivers for the
spread difference between them:
200
0
Jan 98
Jul 98
Feb 99
Aug 99
Mar
Jan 28
0000
Figure 2
MLHY: Market spreads vs. rock-bottom spreads in optimistic

and pessimistic scenarios
Basis points
1200
1000
800
786
600
400
451
307
200
1. EMBI px+ credit quality is higher. The shares of BB

and B-rated issues are roughly the same, but the CCC
component is practically absent and replaced by BBBrated issuers (Figure 3). In addition, the EMBI px+ has
lower market-value-weighted coupon rate (8.46% vs.
9.32%) and longer market-value-weighted maturity (10.1
yrs vs. 7.9 yrs ) than the MLHY. These differences
contribute to a lower rock-bottom spread for the
EMBI px+ .
Figure 3
Credit rating compositions of MLHY and EMBI p+x

Percent
EMBI+px
60
MLHY
50
0
Nov 86 Jan 89 M ar 91 May 93 Aug 95 Oct 97 Dec 99 Mar 02
Which asset class is better depends on the combination

of overall risks and returns each provides. Expected
returns are driven by current spreads and credit
fundamentals. In the absence of spread moves, expected
returns from credit fundamentals for EMBI px+ are 3.8%
over Treasuries, while for MLHY they are 0.4%.
However, these higher emerging markets returns must be
set against the high volatility emerging market investors
40
30
20
10
0
BBB
BB
CCC
Source: Merrill Lynch and J.P. Morgan analytics.

Note: Nigeria is not rated by S&P or Moodys, we assign CCC rating to it based on
its economic fundamentals
New York
October 25, 2001

Portfolio Research
2. MLHY diversity is greater. There are 15 countries in

the EMBI px+ . Latin America predominates, with Argentina,
Brazil, and Mexico constituting 68% of the index. The
other regions Eastern Europe, Asia, and Africa
account for about 20% of total capitalization. In addition,
Moodys regards the risk of simultaneous defaults to be
highest in Latin America. We apply their framework for
measuring concentration risk in the emerging markets
corporate bond sector2 and arrive at a diversity score of 5
for EMBI px+ . In other words, EMBI px+ effectively contains
five independent sources of credit risk, so we will treat a
single default event as affecting 20% of the index. A
countrys market weight, its geographic concentration,
and its default correlation all contribute to the diversity
score.
In contrast, the diversity score for the MLHY is 70.
Currently, the MLHY encompasses 993 issues spanning
43 industries. The largest three industry sectors comprise
30% of the index. While companies in the same industry
are vulnerable to simultaneous default, it is unlikely for
20% of the index to default at the same time. The worst
year was 1991, when 13% of speculative-grade issuers
rated by Moodys defaulted. In other words, diversity in
the high yield market tends to smooth out the incidence of
defaults over time.

page 71
summarized in Table 2 and displayed in full in our earlier

publication.
Table 2
Credit quality transition probabilities

Percent of rated issuers per year
Corporates
Upgrades
Long-term
Downgrades
Defaults
Upgrades
Next two years

Downgrades
Defaults
11.3
0.00
12.0
0.00
AA
1.1
10.2
0.01
0.7
11.5
0.01
3.0
6.8
0.01
3.0
8.6
0.01
BBB
7.5
7.2
0.16
4.9
8.8
0.16
BB
6.4
8.6
1.50
3.0
11.3
2.10
AAA
7.5
2.8
7.09
3.8
4.8
9.49
11.0
26.15
6.0
33.15
Upgrades
Long-term
Downgrades
Defaults
Upgrades
Next two years

Downgrades
Defaults
2.8
0.00
2.4
0.00
AA
0.7
3.0
0.00
1.0
2.2
0.00
4.8
4.4
0.23
4.8
3.3
0.21
BBB
5.2
8.3
0.30
6.8
7.4
0.23
BB
6.5
7.3
1.82
8.6
5.7
1.44
17.2
2.5
5.43
19.5
1.3
4.03
9.8
13.00
12.8
9.06
CCC
Sovereigns
AAA
B
CCC
Table 3
Estimated historical sovereign transition matrix

AAA
3. Emerging market recovery rates are lower. Corporate

bond recovery values have averaged around $453 per $100
of par over the past two decades, with a standard deviation of $23. These estimates, which we use in our spread
calculations are drawn from the experience of some 700
defaults. In contrast, there is little in the way of emerging
market experience to guide us. Only eleven rated or
unrated sovereign issuers have defaulted on foreign
currency bonds since 1975. Many of these defaults were
small-scale. In some cases, the sovereigns did not default
on other foreign currency debt, and so they are of dubious
relevance to EMBI px+ bonds, which are effectively senior
unsecured debt. Currently, Russian Prins and IANs are
trading at $16 and $18, respectively. Ecuador PDIs are
trading at $23. We use these figures as a guide, and
assume for emerging markets a $17.50 average recovery
value, with a $7 standard deviation. Our earlier studies
also show that trading levels of emerging market bonds
imply such recovery values4,5.
4. Emerging markets have better credit migration and
default probabilities than high yield markets. Information
abounds on the credit migration and default experience of
corporate issuers. We use as our baseline estimate the
long-term probabilities calculated by Moodys. These are
AAA
AA
BBB
BB
CCC
Default
97.2
2.8
0.0
0.0
0.0
0.0
0.0
0.0
AA
0.7
96.2
1.6
0.2
0.6
0.7
0.0
0.0
0.0
4.8
90.6
4.4
0.0
0.0
0.0
0.2
BBB
0.0
0.0
5.2
86.1
6.8
1.6
0.0
0.3
BB
0.0
0.0
0.0
6.5
84.4
5.9
1.4
1.8
0.0
0.0
0.0
0.2
17.0
75.0
2.5
5.4
CCC
0.0
0.0
0.0
0.0
0.0
9.8
77.3
13.0
Source: J.P. Morgan estimates.
Once again, there is very little emerging market experience

on which to draw. One fundamental problem derives
from the small number of issuers rated, which imparts a
high degree of uncertainty to any extrapolation of historical estimates. Until 1990, Venezuela was the only
noninvestment-grade sovereign rated by S&P. In the last
five years, S&P assigned significantly more sovereign
ratings (there are now 80), of which one-third are speculative grade.
A second problem derives from the prevalence, until
recently, of loan financing. Because of their less public
nature, loans are more default-prone than bonds. S&P
calculates that since 1980, defaults on foreign currency
bank debt have been at least nine times more frequent than
New York
October 25, 2001

Portfolio Research
those on foreign currency bonds. Extrapolating past loan

experience is thus likely to overestimate future emerging
market defaults. Yet, as bond financing becomes more
prevalent, loans will offer less of a cushion, and so bond
default rates should exceed historical figures. With these
issues in mind, we draw on two sources:

page 72
Figure 4
Sovereign rating action is currently on an upward trend

Percent
30
25
Activity
20
15
S&Ps study of sovereign credit migration focuses on

substantial defaults on foreign currency bond obligations
during the period 1975-19986. The limited size of the
sample is evident in the fact that no CCC-rated sovereign
defaulted. It includes only defaults by Pakistan and Russia.
(Ecuador, which defaulted in 1999, is not rated by S&P
although it is currently rated Caa2 by Moodys, while
Venezuelas default involved bonds primarily held by
domestic residents.)
J.P. Morgans Corporate Risk Management Group (JPMCRMG) has assigned credit ratings to 95 sovereign issuers,
including many emerging market countries based on the
issuers economic fundamentals. JPM-CRMG has also
extended the credit rating history of the sovereigns rated by
S&P. This enlarged rating history includes defaults of both
bonds and loans.
The best baseline estimate of credit migration probabilities

should thus lie somewhere between these extremes, and
so we use a weighted average. In view of the 9:1
historical ratio of loan defaults to bond defaults, it is
unrealistic to lean too close to the JPM-CRMG figures.
We assign a 75% weight to the S&P (bond-based)
transition matrix, and a 25% weight to the JPM-CRMG
matrix for rating categories other than CCC. The
absence of CCC issuers in the S&P sample means we
have to rely on the JPM-CRMG (loan) numbers, whose
default rate we halve and adjust the other credit migration
probabilities accordingly. The resulting default rates
display some broad features of sovereign exposure high
credit ratings are less of a safeguard against defaults (for
example, the A-rated default probability is 0.23%,
compared to Moodys 0.01% for corporates), but CCCrated sovereigns are less likely to default than corporates
(see Tables 2 & 3).
10
Drift
5
0
-5
-10
Dec 90
Oct 92
M ay 96
M ar 98
Dec 99
Source: J.P. Morgan estimates.
rating actions have been on an upward trend recently.

Figure 4 shows rating activity (upgrades plus downgrades) and rating drift (upgrades minus downgrades) as
a percentage of rated sovereigns, according to S&P.
Clearly, rating drift has bottomed out and we expect it to
improve further. We forecast that, over the next two
years, rating drift will be 7%, while activity remains at
current levels. The resulting adjustments to the transition
matrices for the next two years are shown in Table 2.
Calculating Rock-Bottom Spreads
Table 4 summarizes the salient differences between the
two indices and shows that they result in a rock-bottom
Table 4
Summary of assumptions for rock-bottom spreads of MLHY

and EMBI+px
Customized
EMBI+EMBI p+x
MLHY
Credit quality
and composition (%)
BBB
0.0
6.8
BB
40.7
38.5
51.0
54.0
CCC
Coupon (%)
Maturity (yrs)
5. Emerging markets are on an upward credit trend,

whereas high yield markets are on a downward trend.
The transition matrices estimated in the previous paragraph represent credit migration probabilities over the
long term. These averages mask the cyclical dimension
of rating changes and defaults. The speculative-grade
corporate default rate has risen above its historical
average in 1999, and we factor a two-year continuation
of the current level into the transition matrix we use to
calculate rock-bottom spreads1. In contrast, sovereign
Aug 94
Diversity score
8.3
0.6
9.32
8.46
7.9
10.1
70
Recovery value average

(per $100 face)
$45
$17.50
Long-term default rates

6.4
3.7
(credit cycle for next 2 yrs)
2.0
-0.9
Rock-bottom spread (bps)

Market spread (bps)
533
451
604
587
(Index average %)
default rate adjustment (%)
Source: J.P. Morgan estimates and Merrill Lynch.
New York
October 25, 2001

Portfolio Research
Figure 5
MLHY
533bp
Better EM
Composition
-114bp
Lower EM Diversity
+214bp
Lower EM
Recovery
+332bp
Safer EM Credit
Migration
-205bp
Better EM Credit
Outlook
-156bp
+
EMBI px
604bp
spread of 604 bps for EMBI px+ and 533 bps for MLHY.
The contribution of each of the five factors to the 71 bps
spread difference is depicted in Figure 5. Starting with the
makeup of the MLHY, we first calculate the rock-bottom
spread of a hypothetical portfolio that has the credit quality
characteristics of the EMBI px+ but otherwise has the same
characteristics (diversity, credit migration probabilities, etc)
as the MLHY. This would only warrant a rock-bottom
spread of 419 bps, or 114 bps lower than MLHY, because
of the higher credit quality of the EMBI px+ . Next, we lower
the diversity score of this hypothetical portfolio from the
MLHYs 70 to the 5 of the EMBI px+ , retaining all other
characteristics of the first hypothetical portfolio. This
change warrants a rock-bottom spread of an additional 214
bps or 633 bps. We continue in this way until we reach
the full set of characteristics of the EMBI px+ , whose rockbottom spread is 604 bps. It is striking that none of the
five adjustments requires a change in spreads of less than
100 bps. This underpins the importance of a careful
assessment of credit fundamentals. With such large
adjustments in both positive and negative directions,
valuation that does not consider credit fundamentals
explicitly is less than credible.
Our principal rock-bottom spread estimates include our
views on the trend of credit quality in each market. We
can also calculate a long-term rock-bottom spread that

page 73
assumes credit fundamentals will conform to long-term

average conditions. For MLHY, this long-term rockbottom spread is 451 bps, compared to a market average
of 447 bps since the beginning of 1987. For EMBI px+ , the
long-term rock-bottom spread is 661 bps. The average
spread of EMBI px+ over the last two years is 733 bps.
Spread ranges
Spreads paid by the market at any point in time embody a
view of subsequent credit fundamentals. When spreads
are at their highest, this view is at its most pessimistic,
while the lowest market spreads should correspond to
extreme optimism. We have two separate measures of
optimism and pessimism the extremes market spreads
have reached, and the spreads implied by typical extreme
views. In this section, we provide a characterization of
extreme views for each market, and compare them to the
extremes achieved by market spreads.
Historically, when default rates rise above historical
average, recovery values have tended to become lower
than historical average. Therefore, our hypothetical
pessimistic and optimistic scenarios modify both the
transition matrix and recovery values. Table 5 summarizes the default rates involved, and the resulting
rock-bottom spreads are compared with market spreads in
Figures 1 and 2 (page 2).
Table 5
Default rates

Corporates
Optimistic
Long-term
Pessimistic
AA
0.01
0.01
0.01
0.01
0.01
0.01
BBB
0.20
AAA
0.20
0.16
BB
0.3
1.5
3.6
2.3
7.1
15.5
8.0
26.2
50.6
1.95
6.4
13.6
CCC
MLHY
Sovereigns
Optimistic
AAA
Long-term
Pessimistic
AA
0.00
0.00
0.21
0.2
0.90
BBB
0.23
0.3
1.20
BB
1.4
1.8
2.8
4.0
5.4
6.7
CCC
9.1
13.0
26.0
EMBI+px
2.8
3.7
4.9
New York
October 25, 2001

Portfolio Research

page 74
Our view is that optimism and pessimism in emerging

markets are best described by scenarios that involve a
permanent structural change from the current credit
fundamentals environment. The optimistic view is that
emerging market economies will converge to the characteristics of G-7 countries, experiencing persistent rating
upgrades. We characterize this by the view that credit
fundamentals follow our credit cycle upswing scenario
of the preceding section, indefinitely in the future (as
opposed to just the next two years). We also assume an
average recovery value range of $25 per $100. This view
results in a rock-bottom spread of 453 bps. The lowest
level we have seen since the beginning of 1998 for the
EMBI px+ is 380 bps in March 1998.
It is significant that our extreme spreads are based on

extremes of credit fundamentals, rather than the range in
which the market spread has traded. We are not saying
that spreads will tend back to their historical average. We
are not appealing to a mean-reversion model based on
historical spread movements. Rather, our pessimistic
scenarios identify spread levels above which long credit
exposure will pay off whether spreads move or not
because credit fundamentals are likely to be better than
those priced in. Similarly, if the market trades at or
below our optimistic scenario spreads, then long credit
exposure will not pay off, even if spreads do not rise,
because credit fundamentals are likely to be worse than
those priced in.
Our pessimistic view is diametrically opposed emerging

market countries will never free themselves from
indebtedness, and will be subject to persistent default and
restructuring. To characterize this view, we use the
transition matrix estimated by JPM-CRMG, which counts
all defaults, including those on loans, and draws from the
last twenty years, when loan defaults were common. We
reduce the average recovery value to $12.50. These
assumptions lead to a spread level of 1103 bps. As can
be seen in Figure 1, in the last two years, the EMBI px+
overshot to above 1400 bps during the Russia crisis.
Risks and returns
For the high-yield market, we view the extremes as

cyclical peaks and troughs, rather than permanent
changes in credit fundamentals. It is convenient to
express optimistic and pessimistic views in terms of the
arithmetic difference between our view of credit migration over the next two years and the long term transition
matrix. If we add 3.5 times this difference to the longterm transition matrix, we achieve a speculative-grade
default rate of 13.6%. In 1991, the worst year during the
collapse of Drexel, the default rate was 13%. Our
pessimistic view for high yield is that these crisis conditions last for two years, during which average recoveries
drop to $40, and after which historical credit fundamentals resume. The resulting rock-bottom spread is 786 bps
(see Figure 2).
Up to this point, our valuation framework has dealt only

with credit fundamentals. To assess overall value, we
also need to take into account the liquidity conditions of
each market, and how much is available to compensate
for them. We can think of liquidity conditions as having
two dimensions, both of which contribute to the cost or
uncertainty of converting financial assets into cash. One
is the short term volatility of portfolio value. The other is
the ability of investors to liquidate their portfolios without
disrupting the market, i.e., without causing bid-offer
spreads to widen dramatically.
Historically, the EMBI px+ has fluctuated in value much
more than the MLHY. Annual EMBI px+ volatility has run
at about 16%, while the comparable figure for the MLHY
is about 6.5%. Both of these figures are significantly
higher than the volatility to be expected from their credit
fundamentals (Figure 6), although the difference is
particularly glaring in the case of emerging markets.
There is an important relationship between the depth of
emerging markets and their greater volatility. High yield
Figure 6
Return volatility: Market vs. Credit Fundamentals

Percent per year
Our optimistic high-yield scenario departs from the

historical average transition matrix in the opposite
direction by about two times the current adjustment.
Again, this adjustment applies for two years only. Under
this scenario, speculative-grade bonds are much more
likely to be upgraded than downgraded. The implied
default rate for MLHY 1.95%, which, during the nineties, the market only bettered in 1997. Combined with an
increased recovery value of $47.50, the resulting rockbottom spread is 307 bps.
20
16
12
Market
Credit
Fundam entals
8
4
0
HY
+
EMBI+px
EMBI
px
and emerging markets are the two principal means fixed

income bond investors can use to raise their exposure to
credit risk. They can do this in either a tactical (shortterm) or strategic (long-term) way. Tactical allocations to
high yield would be too costly, if not impossible to carry
off, because bid-offer spreads are high and tend to widen
dramatically in the face of even moderate-sized transactions. By the same token, the liquidity conditions of
emerging markets suit them much better to high-frequency turnover. Any time a flight to quality or a decline
in risk appetite occurs, therefore, emerging market
positions are a prime candidate for the adjustment,
whether the precipitating event occurred in emerging
markets or not.
These same conditions make strategic allocations to
emerging markets less attractive for portfolios whose
performance is assessed by frequent mark-to-market. In
the short term, the threat of volatility overshadows the
return benefits, and so the long-term performance of the
asset class does not get a look in. Any investor who
makes strategic investments but is constrained by shortterm volatility will tend to favor high yield over emerging
markets. Yet, the appropriate investment horizon for a
strategic investor is the long term, and, over the long
term, only the credit fundamental volatility of the asset
class remains. As Figure 6 shows, emerging markets and
high yield credit fundamental volatilities are much closer
than their market volatilities. As a result, the long-term
comparison is not so favorable to high yield.
To attach some numbers to these ideas, consider investing
$1 in 10-year zero coupon portfolios either in emerging
markets or high yield. These (necessarily hypothetical)
portfolios have the same credit fundamental characteristics as we have used for their respective markets. Figure
7 charts the mark-to-market value of the high-yield
portfolio as the investment horizon lengthens, or, equivalently, as the bonds tend to maturity. At each investment
horizon, the thick dashed line charts the average growth
of the value of the portfolio. The shaded area indexed the
volatility of the portfolio 90% of the time, the value of
the portfolio will lie within it. As the horizon lengthens
(from zero) its height first increases, reflecting the impact
of market volatility. However, after several years, the gap
starts to narrow, as the effect of the bonds pull-to-par
appears. Were we examining default-free bonds, the
shaded region would converge to a single point at maturity, when the principal would be repaid with certainty. In
the present case, there is a range of uncertainty of the
portfolio value at maturity, because the bonds are subject
to default, and the extent of these defaults is uncertain.
Thus, under our credit fundamental assumptions, the

page 75
Figure 7
10-year zero coupon portfolio over alternative time

horizons: High Yield
Portfolio value

Portfolio Research
3.0
3.0
2.5
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
0
5
6
Horizo n (yrs)
10
average return over the life of a 10-year zero with the

same credit fundamentals as the high-yield market is 97%.
90% of the time, the actual return would be between 83%
and 112%, depending on the default experience that
materialized.
Figure 8 superimposes the performance of the emerging
market portfolio on Figure 7. The (thick, solid) average
total return line is steeper for emerging markets, because
their greater yield, and a lower default rate than high yield
dominate the effect of their lower recovery rate. At short
time horizons, this superior average return appears
insignificant in relation to the effect of market volatility,
represented by the ballooning of the two thing solid lines.
In spite of their higher average return, there is at best only
a 60% chance that emerging markets will outperform high
yield at a one-year time horizon. As with high yield, at
longer investment horizons the pull-to-par effect dominates. At maturity, volatility has declined by about
one-fifth from its peak at seven years, and only the range
of uncertainty caused by defaults remains. This range is
wider than high yields range, largely due to emerging
markets significantly lower diversity score. At worst,
Figure 8
10-year zero coupon portfolio value over alternative time

horizons: High Yield and Emerging Markets
3.5
3.5
High yield
Emerging Markets
3.0
Portfolio value
New York
October 25, 2001
3.0
2.5
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
0
Horizon (yrs)
10
New York
October 25, 2001

Portfolio Research
there is a 78% chance that emerging markets will outperform high yield at a 10-year horizon.
The moral of this story is that emerging market investors
may have a bumpier ride in terms of mark-to-market
volatility, but, according to our credit fundamental
assumptions, superior returns are available to those who
can see beyond this. Investors who can synchronize the
timing of their liability cashflows with those provided by
emerging market sovereign bonds stand to meet those
cashflows with a greater surplus in hand than investors
who operate the same strategy with high yield bonds.
Conclusion
Our credit fundamentals framework allows us to
compare the value of disparate asset classes such as high
yield corporates and emerging market sovereigns. Given
our current views of credit fundamentals, we estimate
rock-bottom spreads of 604 bps and 533 bps for
emerging markets and high yield, respectively. Both of
these spreads exceed current market spreads, indicating
that investors can expect an information ratio of a little
less than one-half in each case.
An advantage of our framework is that it draws on
information that is unaffected by market valuations. This
makes it possible to estimate ranges for market spreads
independently of market conditions. Our conclusion is
that both markets are about 120150 bps above the levels
associated with optimistic credit fundamentals. Given
current views on the direction of credit fundamentals,
(positive for emerging markets, negative for high yield)
this is most disturbing for high yield. We can also isolate
the volatility associated with credit fundamentals, which
we identify as more relevant to strategic investors than
market volatility. The long-term (credit fundamentals)
risk-return tradeoff is more favorable to emerging
markets than the short-term tradeoff, based on market
volatility.

page 76
The theme that runs through our discussion of credit

fundamentals is that a lot of moving parts are involved.
Changes in recovery rates, upgrade or downgrade
frequencies, or diversity all influence the spread
warranted by emerging markets and high yield.
Moreover, since these asset classes embody large credit
risks, movements in these parts cause significant
movements in rock-bottom spreads. This fact of life
underpins the importance of using our framework to
measure value one cannot otherwise hope to keep track
of credit fundamentals exposure.
References
1.
2.
3.
4.
5.
6.
Valuing Credit Fundamentals: Rock-Bottom Spreads,

November 17, 1999, J.P. Morgan Credit Strategy Research.
Emerging Market Collateralized Bond Obligations: An
Overview, Moodys Investors Service, October 25, 1996
Historical Default Rates of Corporate Bond Issuers,
19201998, Moodys Investors Service, January 1999,
and Recoveries on Defaulted Bonds Tied to Seniority
Rankings, Standard & Poors, August 1998
Emerging Markets Debt Recovery Values and Their
Effect on Relative Value Analysis, June 20, 1997, J.P.
Morgan Emerging Markets Research.
Relative Recovery Values of Collateralized and
Uncollateralized Brady Bonds, July 26, 1999, J.P.
Morgan Emerging Markets Research.
Sovereign ratings display stability over two decades,
Standard & Poors, May 1999
New York
October 25, 2001

Portfolio Research
David Xu

page 77
October 13, 2001
Portfolio Research
Emerging Markets versus High Yield: Credit

Fundamentals Revisited
To compensate for credit risk:
- EMBI Global investors need a sovereign spread
of 490bp, which is 200bp below market
- High Yield corporate investors need:
- 300bp for BB credits: 170bp below market
- 750bp for B credits: 20bp above market
These valuations are forward-looking, and embody

the rating agencies sovereign credit outlooks and
their forecasts of speculative grade defaults
The gap between emerging markets and high yield
spreads is a bad indicator of relative value, because
sovereign and corporate credit risks are so
different. Our framework values the sources of
credit risk directly
Introduction
In this note, we determine the spread needed to
compensate investors for bearing the credit risk, as
implied by credit ratings, of Emerging Markets (EM)
sovereigns and high-yield US corporate (HY). At the
start of the year, we concluded that the two markets
were roughly fairly valued1*. Now, however, large gaps
have opened between the markets spreads and their
underlying credit exposures.
We find that EM market spreads exceed the spreads
warranted by credit risk by about 200bp. Most EMBI
Global countries are cheap from this standpoint. The
story for HY is mixed. While the aggregate index is
undervalued by 60bp, BB-rated names are much more so
(by 170bp), and Bs are overvalued to the tune of 20bp.
Our conclusions follow from J.P. Morgans rockbottom spread framework. An assets rock-bottom
spread is the lowest an investor can accept for bearing its
credit exposure. Any asset trading below its rock-bottom
spread will not deliver enough return for its credit risk.
Rock-bottom spreads are driven by credit
fundamentals. An issuers current credit rating implies
the probability that one of its bonds will default at any
All valuations and Exhibits in this publication are based on
information as of October 12, 2000
* Numbers refer to references at the end of the paper
Exhibit 1
Converging EM and HY spreads

(bps)
1,400
1,200
EMBI Global
sovereign spread
1,000
800
600
Merrill Lynch high yield

spread (over 10y UST)
400
200
Jan 99
Jun 99
Nov 99
Apr 00
Sep 00
point during its life. The higher the default probability,

and the lower the ensuing recovery value, the higher will
be the rock-bottom spread. Rock-bottom spreads also
account for investors ability to diversify credit risks
across issuers. Thus EM rock-bottom spreads are higher
for the lack of diversity faced by dedicated EM sovereign
investors. Market spreads figure nowhere in the
calculation of rock-bottom spreads, which thus provide
us with a completely independent valuation.
Investors are currently questioning whether EM spreads
can continue to decline, now that they have traded
through the level of HY spreads (Exhibit 1). Our analysis
shows that, as far as credit fundamentals are concerned,
there is no reason to regard HY spread as a floor for EM.
The two markets credit fundamentals are very different,
and EM is very far above its rock-bottom spread based
on current views, while single-B HY spreads fall short of
rock bottom (Exhibit 2).
Exhibit 2
conceal big differences in credit fundamentals

Market spread minus rock-bottom spread (bps)
250
EM
200
EM BB
EM B
HY BB
150
100
HY
50
0
-50
HY B
New York
October 25, 2001

Portfolio Research

page 78
Market spreads are forward-looking: they embody a view

of future credit conditions. Rock-bottom spreads do the
same, and are calculated from explicit views of how
credit fundamentals will evolve over the life of the
instruments being valued. Thus, rock-bottom spreads
for HY corporate are driven by widely anticipated higherthan-average default rates.
large weight.
Similarly, the probabilities of upgrade, downgrade and

default we assign to sovereigns are driven not only by
their rating, but also by the credit outlook assigned to
them by the rating agencies. The positive current tone of
these outlooks has an important influence on rockbottom spreads. For example, B-rated sovereign issuers
with a positive outlook from S&P have been upgraded
10% more frequently each year than Bs with a stable
outlook. For the case of Brazil, a move from stable to
positive outlook implies a fall in its rock-bottom spread
of 80bp. This translates into a large decline in the rockbottom spread of the EMBI Global, in which Brazil has a
The remainder of this research piece goes through our

reasoning on EM and HY credit valuation in detail. We
provide only a brief summary of how rock-bottom
spreads are calculated (see the Box on this page),
referring the reader to the original research paper for a
full exposition2. The next two sections show how the
agencies outlooks and default views translate into rockbottom spreads. Since differences in rock-bottom
spreads derive from differences in credit fundamentals, in
the final section, we calculate how much each component
of credit fundamentals contributes to the difference
between EM and HY rock-bottom spreads.
Rock-bottom spreads: An overview
adjustment for risk, a bonds rock-bottom spreads

exceeds its breakeven spread (at which a zero return is
earned from taking credit exposure).
A bonds rock-bottom spread prices its potential losses

from default, which we summarize under the name
credit fundamentals. The components of credit
fundamentals are shown in the diagram below. The most
obvious source of loss is an immediate default. The
higher the bonds short-term likelihood of a default and
the lower the amount recovered in default, the higher the
rock-bottom spread. In addition, its credit quality may
change, which in turn alters its probability of defaulting
in subsequent years. As a result, the higher the bonds
chance of a decline in credit quality (which exposes it to
higher default rates in subsequent years), the higher will
be its rock-bottom spread.
Probability of
upgrade,
downgr ade,
no change
Probability
of default
Portfolio
diversity
R eco very rat e
Credit
fundamentals
exposure
It does not make sense to take risk via credit, if credit

cannot deliver a return that is competitive with other
ways of taking risk. Rock-bottom spreads embody this
idea; they indicate the spread necessary for the credit
bonds in question to deliver a risk-adjusted return
comparable with other asset classes. Because of this
It is important to note that rock-bottom spread only

values credit risk. It does not account for any premium
market assigns to liquidity, value of convexity, etc. As
such, rock-bottom spread should not be simplistically
viewed as the spread at which market trades.
The risk involved in holding credit has several

dimensions. Apart from the pure uncertainty of
cashflows (will the bond pay principal plus coupon, or
will it only pay its recovery value?), there is the extent to
which these risks can be mitigated by diversification.
The uncertainty associated with a portfolio of
independent credits is smaller than that associated with a
single credit, consequently, rock-bottom spreads decline
as credit diversity increases. The scope for
diversification within EM is much less than in the US
corporate market, and so the rock-bottom spread for the
same bond held in a dedicated EM sovereign portfolio
would be much higher.
It is useful, as a baseline, to calculate rock-bottom
spreads using historical averages. Thus, we use the
credit migration frequencies over the last 20 years to
represent the probabilities of changes in credit quality and
default, and refer to historical estimates of recovery
rates. However, it is worth mentioning that this is just a
point of departure. Rock-bottom spreads are essentially
forward-looking, and so embody views of future credit
fundamentals. In the body of this paper, we show how
to modify the historical baseline assumptions to produce
figures that represent forward-looking views.
Detailed exposition of rock-bottom spread is contained in
reference 2.
New York
October 25, 2001

Portfolio Research
Emerging Markets Credit Fundamentals
Exhibit 4
Upgrades of emerging markets sovereign issuers have

picked up recently (Exhibit 3). In the past year, Mexicos
Moodys rating and South Africas S&P rating have risen
to investment grade. S&P has also upgraded Mexico,
Bulgaria, Hungary, Poland, and Turkey. In the next three
months we expect Mexicos upgrade to investment grade
by S&P, and Brazil to be upgraded by both agencies. The
current strong positive tone is evident from Exhibit 4,
which shows that countries with a positive outlook
outnumber those with a negative outlook by a large
margin
In contrast to the high yield corporate market, a few
issuers account for a large share of emerging market
indices. Changes in these countries spreads typically
associated with changes in rating have a significant
impact on overall index spreads. And the historical
incidence of rating changes differs dramatically,
depending on the countrys outlook. For example, twothirds of B-rated issuers have been upgraded in the year
after they were assigned a positive outlook. For stable Bissuers, the frequency is only one-third. So, countries
with different outlooks present very different credit risks,
and their valuation needs to take account of this. To
price credit fundamentals accurately, we need to factor in
not only the current rating, but also the current credit
outlook.
To capture the information in rating outlooks, we used
the history of S&P sovereign rating changes and outlooks
to calculate separate probabilities of changes in ratings for
countries with positive, stable, and negative outlooks,
respectively. We then combine this information with the
ratings transition probabilities we estimated for
Ratings outlook of major emerging markets

sovereigns are predominantly positive
S&P
Moodys
Postive
Negative
Postive
Negative
Brazil
Colombia
Brazil
Croatia
Costa Rica
Croatia
Costa Rica
Pakistan
Greece
Egypt
Hungary
Ukraine
Hungary
Paraguay
India
Vietnam
Iceland
Peru
Indonesia
Lebanon
India
Kazakhstan
Korea
Kuwait
Malaysia
Malaysia
Mexico
Russia
Turkey
Slovenia
South Africa
Tunisia
Turkey
Source: S&P and Moodys, as of October, 2000
sovereigns in our earlier paper1, to produce separate

transition probabilities by outlook. Exhibit A1 in the
Appendix shows these transition matrices, and details
their calculation. We assume that positive outlook issuers
are subject to the positive outlook transition matrix for the
coming year, after which their credit prospects are
described by the historical average matrix. This
corresponds to the view that the countrys outlook will
change to stable after one year, either because it has
been upgraded, or because it has failed to meet the
conditions that make it especially likely to be upgraded.
This is a pattern we have observed in S&Ps outlook and
rating history.
This approach expresses credit views much more
precisely than the method we used in our earlier work.
There, we used a more optimistic transition matrix than
the historical one to represent all issuers in the coming
year. Here, only certain issuers qualify for an optimistic
view, while some others warrant a negative view.
Exhibit 3
Sovereign ratings continue drifting upwards

Ratio of number of upgrades to number of downgrades
9
8
7
6
5
4
3
2
1
0
Apr 01
Dec 99
Jul 98
Mar 97
Oct 95
Jun 94
Jan 93
Sep 91
May 90
Source: S&P

page 79
The positive outlook transition matrix clearly has a more

positive tone than the stable transition matrix, whose tone
is, in turn, more positive than the negative outlook matrix
(Exhibits 5, 6). For any rating category, the probability
of an upgrade for positive outlook issuers is higher than
that for stable or negative outlook issuers, and the
probability of a default is lower. This, in turn translates
into sizeable differences in rock-bottom spreads, as
New York
October 25, 2001

Portfolio Research
Exhibit 7 demonstrates. The spread for a positive-outlook

single-B is 100bp narrower than for one with a stable
outlook. This difference comes both from an increased
chance of an upgrade, and from a lower probability of
default over the period when the positive upgrade is in
force.
Exhibit 5
Outlook influences the probability of upgrade

Probability of one notch upgrade by rating and outlook
40%
35%
Negative
Stable
Positive
Recovery rates for EM sovereigns are difficult to

estimate, because public defaults have been rare.
Recently, Ecuador PDIs, Russia 07s, and Ivory Coast
PDIs have traded around $20 when their issuers were in
or near default. In this piece, we remain conservative and
leave the recovery value assumption unchanged from our
previous research piece at $17.5.
30%
25%
20%
15%
10%
5%
0%
A
BBB
BB
Exhibit 6
A negative outlook entails highest default probability

Probability of default by rating and outlook
12%
Positive
10%
Stable
Negative
8%
6%
4%
2%
0%
A
BBB
BB
Exhibit 7
Rating outlooks affect sovereign rock-bottom spreads

Rock-bottom spreads for EMBI Global sovereigns by rating
and outlook
800
700

page 80
Positive
Stable
BBB
Negative
600
500
400
300
200
100
0
BB
Our framework translates each countrys credit

fundamentals its rating, outlook, coupon, maturity and
fundamentals that we apply to all emerging markets
diversity scores, recovery rates and their volatility into a
rock-bottom spread. These results are displayed in
Exhibit 8. A country with a split rating, such as Mexico,
will have a different rock-bottom spread according to
Moodys and S&P, as will countries with the same rating
but different outlooks from the two agencies, for
example, Colombia.
We use market weights to aggregate these individual
countries rock-bottom spreads into a spread for the
EMBI Global. According to S&Ps ratings and outlook,
the EMBI Global rock-bottom spread is currently 494bp,
while Moodys is only 12bp different, at 482bp, in spite
of some very large differences between the two at the
country level. Both of these numbers are very far away
from the EMBI Globals market spread of 690bp. Taken
literally, these numbers mean that, after budgeting an
amount of spread necessary to place EM on an equal
footing with other ways of taking risk, there is still some
202bp of spread remaining to compensate for other
exposures created by an EM investment. Or, simply put,
EM pays 202bp over credit fundamentals.
Assigning a positive or negative outlook to a country
obviously involves taking a view on that countrys credit,
just as assigning a rating does. At the country level, it
makes sense to ask the relative value question of whether
these views have been priced into current market spreads.
Countries whose market spreads differ markedly from
their rock-bottom spread are, with certain caveats,
candidates for over- or underweighting. Exhibit 9 repeats
the data in Exhibit 8, averaging the Moodys and S&P
rock-bottom spreads for each country. Countries above
the line are, on this basis alone, potential overweights.
New York
October 25, 2001

Portfolio Research
Those below the line are expensive relative to their credit

fundamentals, and so are potential underweights.
Among the many caveats attached to using this
information, the following seem the most important.
First, a countrys value relative to credit fundamentals
may mask any number of factors that vitiate the obvious
trade. There may be technical market factors that render
its bonds permanently cheap or expensive relative to

page 81
credit fundamentals. The markets may differ with the

rating agencies, and these differences may drive the
differences in market and rock-bottom spreads. In
particular, the market often responds quicker than the
agencies: an apparently cheap country may be one the
agencies have not downgraded or put on watch, but the
market has. Second, we should be careful to draw
inferences about countries with split ratings, especially
those which one agency rates triple-C. For these
Exhibit 8
Most EMBI Global emerging markets countries look attractive relative to their rock-bottom spread
S&P
Rating and
outlook
Rock-bottom
spread
Moodys
Rating and
outlook
Rock-bottom
spread
494
Average
Market
Rock-bottom
spread
spread
482
488
<
690
Argentina
BB
438
B1
604
521
<
722
Bulgaria
B+
561
B2
632
596
<
804
B+
500
B2
555
527
<
735
CCC
875
CCC
875
875
<
2023
<
EMBI Global
Brazil
*
Ivory coast
Chile
A-
123
Baa1
158
141
China
BBB
169
A3
112
140
BB
590
Ba2
435
512
CCC
1633
CCC
1633
1633
B-
783
Caa2
1132
958
<
BBB-
349
Baa3
349
349
<
BBB+
105
Baa1
105
105
Baa2
161
150
B1
921
794
Ba1
317
359
<
519
Baa3
293
309
<
360
Baa3
Colombia
*
Algeria
Ecuador
Croatia
Hungary
Korea
Lebanon
Morocco
Mexico
Malaysia
*
Nigeria
Panama
Peru
Philippines
BBB
140
B+
667
BB
401
746
749
1303
415
119
<
264
267
326
BBB
165
232
199
<
230
CCC
1218
CCC
1218
1218
<
1887
BB+
350
Baa1
177
263
<
471
BB
522
Ba3
474
498
<
674
BB+
370
Ba1
370
370
<
657
182
182
<
269
<
BBB+
CCC
1224
Thailand
BBB-
259
<
BB+
Poland
**
Russia
182
B+
517
Ukraine
CCC
1690
Venezuela
650
South Africa
BBB-
283
Turkey
213
138
Baa1
B3
770
997
Baa3
259
259
B1
517
517
Caa1
1455
1572
B2
650
650
<
822
Baa3
251
267
<
364
As of October, 2000
As of market close of Oct 12, 2000
* These sovereigns are neither rated by S&P nor by Moody's. We assign CCC ratings to them with stable outlook
** Russia is rated SD (selective default) by S&P and B3 (with positive outlook) by Moody's. Since Russia has just "cured" its default on external debt,
for the purpose of calculating rock-bottom spread, we adjust its S&P rating to CCC with stable outlook
Source: S&P, Moody's, and J.P. Morgan analytics
1056
161
<
624
1622
New York
October 25, 2001

Portfolio Research

page 82
to have almost fully digested the rating and rating outlook

of Mexico until the sell-off during this week: its market
spread was wider than its rock-bottom spread by only
about 10bp at the end of last week. The sell-off during
the past few days increases this gap to 50bp. However,
a rating upgrade by S&P would move the rock-bottom
spread to about 293bp about 70bp tighter than its
current market spread. Clearly, this amount of tightening
derives purely from credit fundamentals. The complete
elevation of Mexico to the investment grade status may
have other ramifications, such as enlarged investor base,
which could have a more significant impact.
countries, there are wide variations in rock-bottom

spread between the two agencies, and the departure from
fair value may be insignificant in this context.
With these caveats in mind, Exhibit 9 suggests that
Ecuador, Venezuela, Bulgaria, Colombia, Argentina,
Brazil, Peru, Philippines, and Panama look most
attractive. Our current portfolio recommendations
overweight many of these countries. Exceptions are
Bulgaria, where we effectively are not so optimistic as
S&P, and Peru, where political uncertainties have arisen
very recently. In this way, the rock-bottom spread
framework can be used as a quick sanity check on less
formal portfolio decisions.
Given the high and volatile level of EM spreads, it is
perhaps remarkable how close market and rock-bottom
are for many countries. For example, the market seems
Exhibit 9
Most EMBI Global emerging market countries look attractive relative to their rock-bottom spread
EMBI Global country market spread (bps) vs rock bottom spread (bps)
1400
Ecuador
1200
Cheap
Russia
1000
Bulgaria
800
Colombia
Brazil
Peru
Philippines
Croatia
Mexico
South Africa
Chile
200
Expensive
Morocco
Panama
Korea
Argentina
Turkey
600
400
Venezuela
Poland
Malaysia
Lebanon
Thailand
China
Hungary
0
0
As of close of Oct 12, 2000
200
400
600
800
1000
1200
1400
B
15%
BB
10%
5%
0%
1980
1985
1990
1995
2000
Exhibit 11
Effect of 1% increase in speculative default rate

Changes of transition matrix (%)
AAA
AA
BBB
BB
CCC
AAA
-0.3
0.3
0.0
0.0
0.0
0.0
0.0
0.0
AA
-0.2
-0.5
0.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-0.9
0.8
0.1
0.0
0.0
0.0
BBB
0.0
-0.1
-1.3
0.5
0.7
0.1
0.0
0.1
BB
0.0
0.0
-0.2
-1.5
0.0
1.2
0.1
0.3
0.0
0.0
0.0
-0.2
-1.7
-0.3
1.0
1.2
CCC
0.0
0.0
-0.2
-0.2
-0.7
-1.4
-1.0
3.5
Exhibit 12
Change in credit migration probabilities accompanying

a one percent increase in speculative grade default
(%)
4
CCC
3
2
B
1
BB
0
Upgrades
(number of notches)
-5
-4
-3
-2
-2
-1
-1
Historically, changes in the default rate have typically

been accompanied by changes in the entire pattern of
speculative grade credit rating changes. When defaults
have been above average, there have tended to be less
upgrades, and more downgrades than the historical
average. In addition, the experience of issuers in all rating
categories is affected similarly. Exhibit 10 provides an
illustration. The correlation between annual B and BB
default rates since 1980 is 0.72. Because of this similarity
among profiles, it makes sense to estimate (from the
historical record) the typical change in the transition
20%
No change
We now assemble the ingredients necessary to calculate

rock-bottom spreads that embody Moodys current view.
Probabilities of upgrade, downgrade and default are
collected together in a credit migration or transition
matrix, as shown in Exhibit A2 in the Appendix. This
matrix contains the average annual number of rating
changes and defaults in each rating category, over the
period 1983-1999.. For example, of the issuers that
started a year rated BB, on average 8% were downgraded
to B during the year. If you have no strong view on the
direction of credit trends, these numbers are a plausible
representation of what you might expect, since they
embody the historical average. However, a view that
default rates will differ from historical averages demands
a different transition matrix.
Default rates are correlated
+1
Moodys also report that the average price of defaulted

senior unsecured bonds one month after default has fallen
to $37 per $100. This figure reflects the price at which
defaulted bonds can currently be disposed of. As the
markets discounted valuation of what will be paid out
when the issuer emerges from default, it is a meaningful
indicator of recovery rates on defaulted bonds. It
compares with a historical average recovery rate of $45.
Exhibit 10
+2
Moodys projects that the rate of default among

speculative issuers will rise to 8.4% over the year ending
July 2001. This compares to a 5.8% rate over the last 12
months, and a historical average of 4.1% per year since
1983. In our earlier study1, we projected a 6% default
rate for each of 2000 and 2001. Here, we take a middle
ground, based on our view that Moodys forecast is in
the right direction, but may be on the high side, especially
given the recent slowdown and minor decline in the
default rate. We shall use a speculative grade default rate
of 7.5%, and explore the implications of it being in place
both over the next year, and over the next two years,
after which default rates revert to historical averages.
matrix to be expected when, for example, there is a 1%

increase in the aggregate speculative grade default rate.
Exhibit 11 details these adjustments. Exhibit 12 makes
them easier to understand, by displaying them in terms of
the number of notches by which the rating changes
over the year. The patterns for speculative grades are
+3
High Yield

page 83
+4

Portfolio Research
+5
New York
October 25, 2001
Downgrades
(number of notches)
New York
October 25, 2001

Portfolio Research
broadly similar, the principal shift involving a 1.5%

decline in the one-notch upgrade probability, and a
concomitant increase in downgrade or default
probabilities.
The recipe for the transition matrix that corresponds to
the view that the default rate over the next year will be
7.5% (i.e., 3.4% above the historical average) is now
quite simple. To each number in the historical matrix
(Exhibit A2), simply add 3.4 times the corresponding
element in Exhibit 11, to produce the matrix in Exhibit
A3.
This calculation characterizes Moodys default view in
terms of credit migration for the next year. At the end of
that year, any bond that has not matured will trade
according to credit prospects during the rest of its life,
and so the view must also stipulate a transition matrix for
each of these following years. As Moodys do not take
default views for time horizons longer than one year, we
look at two default rate scenarios for the second year:
one a repeat of the 7.5% default rate, the other an
immediate reversion to the 4.1% historical average
default rate (this will be our central scenario). For
subsequent years our best guess is that the default rate
has returned to its historical average, and so we use the
historical average transition matrix (Exhibit A2)
Exhibit 13 displays the rock-bottom spreads that result
from these HY credit fundamentals views. The current
spread for the market as a whole, as measured by Merrill
Lynchs High Yield Index, is 60bp above its rock-bottom
spread of 648bp (fourth column of Exhibit 13), under the
assumption that adverse credit conditions persist only for
the next 12 months. However, this undervaluation
relative to credit fundamentals derives solely from the BB
sector, where market spreads exceed rock-bottom by
171bp. Both B-rated and CCC-rated issuers credit
fundamentals are overpriced by 20 or so basis points.
Exhibit 13
Historical market and rock-bottom spread levels

Market
May
Rock-Bottom Spreads
Oct
1999 2000
Historical
2000-01
2000-02
Credit
Fundamentals
Default
Scenario
Default
Scenario
BB
290
467
231
296
475
733.4
572
752
852
CCC
1236
1894
1389
1909
2099
MLHY
465
708
491
648
730
As of close of Oct 12, 2000
336

page 84
Exhibit 14
The BB sector has responded most sharply to rising

defaults
Market spreads (bps) versus rock-bottom spreads (bps)
800
700
600
500
BB
400
300
200
Oct 2000
100
May 1999
0
0
200
400
600
800
The deterioration of the market's view on HY dates to

late 1999, when evidence of increased HY defaults began
to surface. Earlier, say in May 1999, there was
something of a lull in the view of credit fundamentals,
when there seemed to be no strong view in either
direction. So, we assume that at that time, the market's
view conformed to historical credit fundamentals,
resulting in the rock-bottom spreads in the third column
of Exhibit 13. Similarly, we pair current market spreads
with the view that high default rates will persist for one
year (fourth column of Exhibit 13). The changes that
have occurred in views and market spreads are apparent
from Exhibit 14. Both B and BB spreads have risen, as
have rock-bottom spreads. However, while single-B
market spreads have only just kept pace with the rise in
rock-bottom spreads, BB market spreads have risen
considerably more than rock-bottom.
The final column of Exhibit 13 displays rock-bottom
spreads in the event that high default rates persist for the
next two years. Single B spreads rise by 100bp, to a
level that is 120bp above current market spreads.
However, BB spreads rise by only 40bp, which still
leaves substantial cushion relative to current market
spreads.
New York
October 25, 2001

Portfolio Research
Comparing Emerging Markets and High Yield

As the foregoing shows, credit fundamentals have many
moving parts. Differences in credit quality, diversity,
credit outlook, recovery rates all conspire to make the
difference between market spreads a very muddy
indicator of relative credit value. As a consequence, we
find it more useful to compare each markets spread to an
indicator of its credit exposure the rock-bottom spread.
However, our framework enables us to measure the
contribution to rock-bottom spreads of each of the
differences in credit fundamentals.
Exhibit 15 charts a path from the 488bp rock-bottom
spread that we calculate for EM, to the 648bp we
calculate for HY. At each step along the path, we change
a credit fundamental feature, from its value in the EM
context to that in the HY context.
As a first step, we change the EM view from our
scenario, based on the rating agencies current rating
outlooks, to a scenario where all outlooks are assumed to
be stable. In other words, we treat each country as if it
had a stable outlook. This results in only a 26bp rise in
the EMBI Global rock-bottom spread, to 514bp.
However, underlying this relatively small change at the
index level are more significant changes at the individual
country level, as pointed out earlier.
Next, imagine that the credit rating composition of the
EMBI Global changes, so that each rating category is
represented by its proportions in HY Index. Since HY
credit quality is lower, we would expect the resulting
spread to be wider. Indeed, as shown in Exhibit 15,
spread widens by 157bp. Thus, were the EMBI Global to
have the same credit makeup as HY (but otherwise
unchanged), a neutral credit view implies a rock-bottom
spread of 671bp.
The historical recovery rate for HY is substantially higher,
at $45 per $100, compared with the $17.5 we assume for
EM. In contrast, the historical pattern of US corporate
credit migration involves higher default rates for
speculative grade issuers (Exhibit A1 versus Exhibit A2).
As Exhibit 15 shows, incorporating each of these features
in turn causes a swing of about 200bp in rock-bottom
spread, narrowing as we switch to the higher HY
recovery rate, and then widening as we switch to HYs
worse credit migration patterns. Thus, an EMBI Global
portfolio with the same composition, recovery rate and
credit migration as HY, and a neutral view on credit
fundamentals, would occasion a rock-bottom spread of
691bp

page 85
Exhibit 15

EMBI Global with
current rating
outlooks
488bp
with stable
rating outlooks
514bp
+26bp
and with HY
credit composition
671bp
+157bp
and with historical

HY recovery rate
473bp
-198bp
and with historical

HY credit migration
691bp
+218bp
and with HY
diversity
491bp
-200bp
and with current HY

recovery rate view
551bp
+60bp
and with current HY

credit migration view
648bp
+97bp
HY with current
default view
648bp
What other differences are there? Much greater

diversification of credit exposure is possible within a
dedicated HY portfolio than in one dedicated to EM. We
describe this by Moodys diversity score, which
effectively translates the group of correlated credit
exposures in each portfolio into a smaller number of
credits that are assumed uncorrelated3. While the
diversity score for HY is around 70 (i.e., holding the HY
market portfolio gives you the same diversity of credit
exposure as you would achieve if you held 70 genuinely
independent credits, EMs diversity score is only 9.
Adjusting our hypothetical portfolio to the diversity of HY
narrows its spread by 200bp. Evidently, lack of diversity
is a big widening contributor to the EM rock-bottom
New York
October 25, 2001

Portfolio Research
spread. However, a non-dedicated investor in EM does

not face this limitation. For example, a marginal
investment in EM sovereigns by a dedicated HY investor
enjoys the diversity of the HY market (or better), not the
lower diversity of EM considered in isolation. Thus, for
crossover investors, the 200bp charge for EMs lower
diversity is not relevant, and EM is accordingly that
much more attractive to these investors. For them,
pricing diversity as if they were dedicated EM investors
is the same as demanding to be paid for non-systemic
risk in, say, an equity portfolio. Finance theory tells us
the reward the market will offer for bearing nonsystemic risk is zero, because this risk can be eliminated
by diversification. This is precisely what the crossover
investor can achieve by investing small amounts in EM.
At this stage in exhibit 15, we have transformed the
credit fundamentals of EM in a series of steps that have
resulted in the credit fundamentals of the HY index under
the assumption of historical (HY) credit migration and
recovery rates. This warrants a rock-bottom spread of
491bp (see also Exhibit 13). To reach the 648bp rockbottom spread associated with our central scenario, we
need to incorporate two further steps. First, our central
HY scenario assumes a $37 recovery rate, compared
with the $45 historical average, in line with Moodys
latest estimates. This widens the HY rock-bottom spread
by 60bp, to 551bp. (So, a useful rule of thumb is that a
$1 decline in the recovery rate widens the HY rockbottom spread by 7-8bp. Obviously, this rule will not be
reliable if the composition of the index changes
significantly). Second, as we move from the historical
or neutral view on HY default rates to the central
scenario, embodying a 7.5% speculative default rate in
2000-01, the rock-bottom spread widens by a further
97bp, reaching the 648bp we have assigned to HY.
Evidently, substantial swings in rock-bottom spreads
occur as we move through the successive changes in
credit fundamentals depicted in Exhibit 15. Changing
views of credit fundamentals have a similar impact on
valuation. What if you believe the EM recovery rate is
better estimated at $25, than our $17.5 assumption?
(Then the EM rock-bottom spread falls by 53bp). What
if you believe that the HY default rate has reached a
maximum, and will go back to its historical average over
the next year? (Then the HY rock-bottom spread falls by
44bp). What if you view EM from the standpoint of a
HY investor, rather than a dedicated EM investor? (Then,
as we have seen, the rock-bottom spread you associate

page 86
with EM should be 200bp tighter). To assess relative

value, it is thus imperative to pin down credit
fundamentals and value them consistently. Simply
comparing the market spreads of EM and HY aggregate
indices is treacherous. If the rest of the market responds
just to the spread differential in deciding on relative value,
it makes sense. But it has no grounding in credit
fundamentals.
New York
October 25, 2001

Portfolio Research

page 87
Appendix
Sovereign and High Yield Transition Matrices

Exhibit A1
Sovereign transition matrix under Positive, Stable, and Negative outlook

Positive
AAA
AA
BBB
BB
CCC
Default
AAA
1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
AA
0.00
0.99
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.07
0.91
0.02
0.00
0.00
0.00
0.00
BBB
0.00
0.00
0.11
0.87
0.01
0.00
0.00
0.00
BB
0.00
0.00
0.00
0.12
0.87
0.00
0.00
0.01
0.00
0.00
0.00
0.00
0.34
0.62
0.02
0.01
CCC
0.00
0.00
0.00
0.00
0.00
0.10
0.77
0.13
Stable (or Historical)

AAA
AA
BBB
BB
CCC
Default
AAA
0.97
0.03
0.00
0.00
0.00
0.00
0.00
0.00
AA
0.01
0.96
0.02
0.00
0.01
0.01
0.00
0.00
0.00
0.05
0.91
0.04
0.00
0.00
0.00
0.00
BBB
0.00
0.00
0.05
0.86
0.07
0.02
0.00
0.00
BB
0.00
0.00
0.00
0.06
0.84
0.06
0.01
0.02
0.00
0.00
0.00
0.00
0.17
0.75
0.02
0.05
CCC
0.00
0.00
0.00
0.00
0.00
0.10
0.77
0.13
Negative
AAA
AA
BBB
BB
CCC
Default
AAA
0.71
0.29
0.00
0.00
0.00
0.00
0.00
0.00
AA
0.00
0.99
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.01
0.84
0.15
0.00
0.00
0.00
0.00
BBB
0.00
0.00
0.01
0.85
0.01
0.13
0.00
0.00
BB
0.00
0.00
0.00
0.04
0.67
0.17
0.06
0.06
0.00
0.00
0.00
0.00
0.17
0.70
0.02
0.10
CCC
0.00
0.00
0.00
0.00
0.00
0.10
0.77
0.13
Explanation of these sovereign transition matrices:

In order to take rating outlook into account, we examined the history of S&P one-year
ratings migration for countries with positive outlooks, negative outlooks, and irrespective of
outlook (unconditional), respectively.
In our earlier research, we estimated a historical sovereign transition matrix using both loan
defaults and extended sovereign rating history based on J.P.Morgan internal ratings.
This historical estimate thus has a different basis than the raw numbers conditioned on outlook,
which only draw on S&P ratings. In order to account for these differences, we obtain the positive
(negative) outlook transition matrix in this exhibit by adding to this historical transition matrix the
difference between the positive (negative) outlook matrix and the unconditional transition matrix.
The stable transition matrix in this exhibit is simply the historical transition matrix.
New York
October 25, 2001

Portfolio Research

page 88
Exhibit A2
Hisorical US corporate transition matrix

AAA
AA
BBB
BB
CCC
Default
AAA
0.89
0.10
0.01
0.00
0.00
0.00
0.00
0.00
AA
0.01
0.89
0.10
0.00
0.00
0.00
0.00
0.00
0.00
0.03
0.90
0.06
0.01
0.00
0.00
0.00
BBB
0.00
0.00
0.07
0.85
0.06
0.01
0.00
0.00
BB
0.00
0.00
0.01
0.06
0.84
0.08
0.00
0.01
0.00
0.00
0.00
0.01
0.07
0.83
0.03
0.07
CCC
0.00
0.00
0.01
0.01
0.03
0.06
0.63
0.26
Exhibit A3
US Corporate Transition Matrix correpsonding to a 7.5% speculative default rate

AAA
AA
BBB
BB
CCC
Default
AAA
0.89
0.10
0.01
0.00
0.00
0.00
0.00
0.00
AA
0.00
0.87
0.12
0.00
0.00
0.00
0.00
0.00
0.00
0.03
0.88
0.08
0.01
0.00
0.00
0.00
BBB
0.00
0.00
0.03
0.87
0.08
0.01
0.00
0.00
BB
0.00
0.00
0.00
0.01
0.84
0.12
0.01
0.03
0.00
0.00
0.00
0.00
0.01
0.81
0.06
0.11
CCC
0.00
0.00
-0.01
0.00
0.01
0.01
0.58
0.40
References
1. Comparing Credit Fundamentals: Emerging Markets
versus High Yield, February 2, 2000, J.P.Morgan
Portfolio Research and Emerging Markets Research.
2. Valuing Credit Fundamentals: Rock-Bottom Spreads,
November 17, 1999, J.P.Morgan Portfolio Research.
3. Emerging Market Collateralized Bond Obligations:
An Overview, Moodys Investors Service, October 25,
1996.
New York
October 25, 2001
J.P. Morgan Securities Asia Pte. Limited

Fixed Income Research
Tatsushi Kishimoto (81-3) 5573-1521

page 89
June 21, 2001
Credit Research
Market Strategy Report
US credits look attractive for Japanese investors

l
Investors in yen credit products are facing two challenges: excess demand for
credit bonds and poor diversification
We believe that both of these challenges can be met by investing in other credit
markets
The US market comes out on top both on a LIBOR basis and from a credit
fundamentals perspective
Investors in yen credit products are currently facing two challenges.

First, there is excess demand for credit bonds. Prompted in part by the BOJs zero-rate
commitment, investors are rushing into the corporate bond market, trying to find a place
to invest their money. At the same time, supply is limited, as a result of the prolonged
recession and corporates attempts to reduce debt. While excess demand may well keep
spreads from widening in the near future, spreads have nevertheless fallen to levels that
do not compensate adequately for the credit risk that is being taken. The short-run
advantages of corporate bonds that they pay more than JGBs must always be balanced against their long-run disadvantage that they entail credit risk. All else equal, no
matter how low JGB rates fall, corporate spreads should move minimally, because they
compensate for the risk of losing the bonds principal, and this risk has not diminished.
Exhibit 1
Spread movements in the Japanese market (5-year maturity)

Spreads over JGB, bp
140
120
100
Baa
80
60
40
Aa
20
Jan 01
Apr 01
Oct 00
Jul 00
Apr 00
Jan 00
Jul 99
Oct 99
Apr 99
Jan 99
Oct 98
Jul 98
Jan 98
Apr 98
Ratings: Moodys
Source: JPMorgan
Second, the credit risk that Japanese corporate bond investors take is more acute for the
fact that it is difficult to hold a diversified Japanese credit portfolio. Issuance is highly
concentrated: in FY3/2001 utility bonds and bank straight bonds together accounted for
44.5% of new corporate bond offerings. Other sectors including manufacturers,
chemicals, pharmaceuticals, foods and transportation are also sporadically issuing
bonds, but the amounts of issuance are relatively small and far from enough to enable
investors to well diversify their credit portfolios.
New York
October 25, 2001

page 90

We believe that both of these challenges can be met by investing in other credit markets.
To substantiate our view, this report compares value in the Japanese corporate bond
market with the US, where spreads appear wider, and a broader array of industries are
represented.
Of course, it is not sufficient to compare Japanese issuers spreads over JGBs with US
issuers spreads over Treasuries, since the latter comes in the form of US dollar cash
flows. Rather, we indicate explicitly how much spread can be earned by a yen investor
in US corporates, taking into account the costs of translating US dollar cash flows into
yen, via a cross-currency swap.
Even after US opportunities have been translated into yen, it does not follow that the
bond paying the best LIBOR spread is the best deal, because we need to take into
account the credit risk involved in each case. We do this by explicitly valuing the credit
fundamentals of each market, using JPMorgans Rock-Bottom Spread1) valuation
framework, which prices the potential for downgrades and defaults to occur, how
much will be recovered in default, and the extent to which these risks can be diversified.
A bonds rock-bottom spread measures how much you need to be paid to earn a competitive return for the risk you are taking. The correct measure of the value of a bond is
not its raw market spread, but the excess of this market spread over its rock-bottom
spread.
On both counts, the US market comes out on top. In any rating category, asset
swapped spreads are wider in the US than in Japan. Moreover, credit fundamentals are
more favorable in the US, where investment grade bonds pay significantly more than
their rock-bottom spreads. In contrast, Japanese credit bond spreads are too low to
cover their rock-bottom spreads, and so cannot be expected to earn an adequate return
for the credit risk they involve (Exhibit 2).
Exhibit 2
Surplus spreads above rock-bottom spreads (5-year maturity) : Japan vs. US

Spreads over LIBOR, bp
100
US
50
0
-50
-100
Japan
-150
-200
-250
Aa1
Aa2
Aa3
A1
A2
A3
Baa1
Baa2
Baa3
Ba1
Source: JPMorgan
1)
Please see Valuing Credit Fundamentals: Rock-bottom Spreads (Peter Rappoport, November 17,
1999) for the details of the rock-bottom spread framework
New York
October 25, 2001


page 91
Spread comparison between the US market and the Japanese market

In Exhibit 3, we identify some corporate bonds issued by US firms that we consider to
be relatively familiar to Japanese investors, and compared their spreads over governments with those of comparable Japanese names.
Exhibit 3
Government spreads of the US corporate bonds and the Japanese corporate bonds
Indicative bids as of June 14, 2001
Issuer
Ratings
Years
Treasury Similar issuer in Japan

spread
Ratings
Years
JGB
spread
Bank of America Corp
Aa2/A+
3.9
83
Shizuoka Bank
Aa2/AA-
3.9
16
CITIGROUP Inc
Aa2/AA-
4.5
90
Shizuoka Bank
Aa2/AA-
4.3
17
Wal-Mart Stores
Aa2/AA
3.2
57
Ito-Yokado
Aa3/AA+
3.0
11
Bank One Corp
Aa3/A
4.1
100
Shoko Chukin
Aa3/A
4.1
12
Chase Manhattan Corp
Aa3/AA-
3.5
93
Shoko Chukin
Aa3/A
3.5
11
Sony Corporation
Aa3/A+
1.7
70
Sony Corp.
Aa3/A+
2.0
Dow Chemical
A1/A
4.8
100
Asahi Kasei
A2/A-
5.0
15
Goldman Sachs Group
A1/A+
4.2
102
Nomura Securities
Baa1/BBB+
4.0
26
Statoil
A1/AA-
4.9
115
Nippon Mitsubishi Oil
A3/BB+
5.0
21
American Airlines Inc
A2/A
3.3
165
Japan Airlines
Baa3/BB
3.0
47
GMAC
A2/A
4.6
135
Fuji Heavy Ind.
A- (R&I)
5.0
33
DaimlerChrysler NA Hldg
A3/A-
4.6
153
Fuji Heavy Ind.
A- (R&I)
5.0
33
Worldcom Inc.
A3/BBB+
1.9
160
Nippon Telecom
AA (JCR)
2.0
Sun Microsystems Inc*
Baa1/BBB+
3.2
165
NEC Corporation
Baa1/ -
3.0
15
FedEx Corp
Baa2/BBB
2.7
132
Yamato Transport
A3/A-
3.0
15
FedEx Corp
Baa2/BBB
4.7
152
Yamato Transport
A3/A-
5.0
10
Kellogg Co
Baa2/BBB
1.8
107
Suntory
Baa3/ -
2.0
18
Kellogg Co
Baa2/BBB
4.8
127
Suntory
Baa3/ -
5.0
24
Delta Air Lines
Baa3/BBB-
4.5
295
ANA
Baa3/BB-
5.0
66
Lockheed Martin Corp
Baa3/BBB-
4.9
135
IHI
Baa2/BBB-
5.0
24
Raytheon Co
Baa3/BBB-
3.8
185
Kawasaki Heavy Ind.
Baa2/BB+
4.0
31
*Callable bond
Source: JPMorgan
Spreads on US names look far wider than those of Japanese names at first glance.
However, as we already discussed, this is like comparing apples and oranges. Comparisons of spreads over governments denominated in two different currencies will
only be accurate if the government and swap curves, respectively, happen to coincide.
We also cannot know from this table how large an extra risk premium each bond has
over its credit fundamentals.
To solve these problems, we take two steps: firstly we will compare the US bonds and
the Japanese bonds on a LIBOR basis, using cross-currency asset swaps to convert
the cash flow of US bonds into -denominated flows; secondly, we will make a relative
value analysis from a credit fundamentals perspective, using our rock-bottom spread
framework.
New York
October 25, 2001


page 92
Spread comparison on a LIBOR basis

Cross-currency asset swaps can remove currency and relative curve risks2) and leave
only credit risks translated into yen by converting dollar-denominated fixed rate cash
flows of the US domestic corporate bonds into -denominated floating rate cash flows.
(See Appendix 2 on page 16 for the details of the structure.)
Exhibit 4
LIBOR spreads comparison

Issuer
Ratings
Maturity

CITIGROUP Inc
Wal-Mart Stores
Bank One Corp
Sony Corporation
Dow Chemical
Goldman Sachs Group
STATOIL
GMAC
Worldcom Inc.
FedEx Corp
FedEx Corp
Kellogg Co
Kellogg Co
Delta Air Lines
Raytheon Co
Aa2/A+
Aa2/AAAa2/AA
Aa3/A
Aa3/AAAa3/A+
A1/A
A1/A+
A1/AAA2/A
A2/A
A3/AA3/BBB+
Baa1/BBB+
Baa2/BBB
Baa2/BBB
Baa2/BBB
Baa2/BBB
Baa3/BBBBaa3/BBBBaa3/BBB-
5/16/05
12/1/05
8/10/04
8/1/05
12/1/04
3/4/03
4/1/06
8/17/05
5/1/06
10/15/04
1/15/06
1/18/06
5/15/03
8/15/04
2/12/04
2/15/06
4/1/03
4/1/06
12/15/05
5/15/06
3/15/05
Years
3.9
4.5
3.1
4.1
3.5
1.7
4.8
4.2
4.9
3.3
4.6
4.6
1.9
3.2
2.7
4.7
1.8
4.8
4.5
4.9
3.7
LIBOR Similar issuer in Japan

spread
(a)
22
15
18
32
42
18
20
33
29
110
53
68
95
117
35
65
50
39
192
47
114
Shizuoka Bank
Shizuoka Bank
Ito-Yokado
Shoko Chukin
Shoko Chukin
Sony Corp.
Asahi Kasei
Nomura Securities
Japan Airlines
Fuji Heavy Ind.
Fuji Heavy Ind.
Nippon Telecom
NEC Corporation
Yamato Transport
Yamato Transport
Suntory
Suntory
ANA
IHI
Kawasaki Heavy Ind.
Ratings
Aa2/AAAa2/AAAa3/AA+
Aa3/A
Aa3/A
Aa3/A+
A2/ABaa1/BBB+
A3/BB+
Baa3/BB
A- (R&I)
A- (R&I)
AA (JCR)
Baa1/ A3/AA3/ABaa3/ Baa3/ Baa3/BBBaa2/BBBBaa2/BB+
Years
LIBOR
spread
(b)
LIBOR
spread
diff.
(a)-(b)
3.9
4.3
3.0
4.1
3.5
2.0
5.0
4.0
5.0
3.0
5.0
5.0
2.0
3.0
3.0
5.0
2.0
5.0
5.0
5.0
4.0
12
13
4
10
7
0
12
23
18
40
30
30
0
8
8
7
10
21
63
21
28
10
2
14
22
35
18
8
10
11
70
23
38
95
109
27
58
40
18
129
26
86
* Callable bond
Source: JPMorgan
Exhibit 4 compares spreads of the bonds listed in Exhibit 3 on a LIBOR basis. As

LIBOR represents a funding rate for investors, these LIBOR spreads show how much
yen-based investors can earn by holding these bonds. Spreads in the US are still wider
than those in Japan, but the difference is smaller than in Exhibit 3, because of larger
swap spreads in the US.
Relative value analysis from a credit fundamentals perspective
The LIBOR spread comparison showed that the US corporate bonds we look at in this
report are generally cheaper than comparable Japanese names. However, the two
markets are driven by different credit fundamentals. We need to calculate how much
each pays net of its credit fundamentals. We will examine this with our rock-bottom
spread framework.
To be more precise, investors who hold a dollar-denominated bond swapped into yen will take a currency
risk if the underlying bond defaults as the currency swap position will over-hedge the currency exposure of
the principal and the remaining coupons.
2)
New York
October 25, 2001


page 93
While the Japanese economy looks to be on the brink of a deflationary spiral, the US
economy is also significantly slowing down since last year, causing the downgrades of
investment grade issuers and speculative grade defaults to sharply increase. To reflect
this in calculating rock-bottom spreads, we used S&Ps historical rating transition
matrix as a baseline, and assumed that credit conditions in 2001 and 2002 will deviate
from this baseline as follows:
l Probability of downgrades of high grade issuers increases to 11.5% from the historical average 7.5%
l Default probability of speculative grade issuers increases to 8.0% from the historical
average 4.1%
We also calculated rock-bottom spreads both under 45% recovery rates (historical
average in the US) and under the assumption relevant to the latest market conditions
(34% for the US market and 15% for Japan).
Exhibit 5
Market spreads and rock-bottom spreads in the US corporate bond market

250
Market spreads
200
Rock-bottom spread (34% recovery)

150
100
50
0
Aa1
Aa2
Aa3
A1
A2
A3
Baa1
Baa2
Baa3
Ba1
Source: JPMorgan
Exhibit 6
Market spreads and rock-bottom spreads in the Japanese corporate bond market
300
Market spread
250

200
150
100
50
0
Aa1
Aa2
Source: JPMorgan
Aa3
A1
A2
A3
Baa1
Baa2
Baa3
Ba1
New York
October 25, 2001


page 94
Exhibit 7
Surplus spreads above rock-bottom spreads: Japan vs. US

100
US
50
0
-50
-100
Japan
-150
-200
-250
Aa1
Aa2
Aa3
A1
A2
A3
Baa1
Baa2
Baa3
Ba1
Source: JPMorgan
We compared market spreads and rock-bottom spreads in the US corporate bond

market and the Japanese market in Exhibit 5 & 6. We used LIBOR spreads as market
spreads for both markets to make comparison consistent. Exhibit 7 compares the
market spreads surplus above rock-bottom spreads in the US market (under 34%
recovery rate) and the Japanese market (under 15% recovery rate). (All spreads are
for five-year maturity.)
The rock-bottom spread is a minimum spread required for investors to take a credit
risk exposure of corporate bonds3). We consider the market spreads surplus above a
rock-bottom spread as a cushion to compensate for non-credit exposure from holding
corporate bonds such as lower liquidity compared to government bonds. We believe
that such surplus-above-rock-bottom-spread (we sometimes call it liquidity spread
for the sake of convenience) should be positive in normal market conditions, and we
might reasonably say that a bond with a bigger surplus is cheaper than others. In
Exhibit 7, market spreads surplus above rock-bottom spreads in the US market are
larger than those in the Japanese market. Japanese credits look very expensive with
market spreads significantly tight relative to rock-bottom spreads. Moreover, in the
Japanese market, market spreads fall short of rock-bottom spreads in most rating
categories. It means that the investors are not paid enough for taking credit. It is only
in the high yield market (at or below Ba1 category) where we can see a rock-bottom
spread exceeding a market spread in the US market.
3)
A rock-bottom spread over LIBOR is theoretically narrower than a rock-bottom spread over government
as a swap transaction involves a credit risk of a counterparty. However, as a plain vanilla swap does not
involve principal movements, a credit risk of a swap is minimal and, therefore, the difference between a
rock-bottom spread over LIBOR and a rock-bottom spread over government is negligible (1~2 bp for 10
years maturity).
New York
October 25, 2001


page 95
Exhibit 8
Rock-bottom spread analysis for individual names in the US market and the Japanese market
Issuer
Ratings

CITIGROUP Inc
Wal-Mart Stores
Bank One Corp
Sony Corporation
Dow Chemical
Goldman Sachs Group
STATOIL
GMAC
Worldcom Inc.
FedEx Corp
FedEx Corp
Kellogg Co
Kellogg Co
Delta Air Lines
Raytheon Co
Aa2/A+
Aa2/AAAa2/AA
Aa3/A
Aa3/AAAa3/A+
A1/A
A1/A+
A1/AAA2/A
A2/A
A3/AA3/BBB+
Baa1/BBB+
Baa2/BBB
Baa2/BBB
Baa2/BBB
Baa2/BBB
Baa3/BBBBaa3/BBBBaa3/BBB-
Years
3.9
4.5
3.2
4.1
3.5
1.7
4.8
4.2
4.9
3.3
4.6
4.6
1.9
3.2
2.7
4.7
1.8
4.8
4.5
4.9
3.8
LIBOR
spread
(a)
Rockbottom
spread
(b)
22
15
18
32
42
18
20
33
29
110
53
68
95
117
35
65
50
39
192
47
114
1
3
1
4
4
2
6
4
7
5
9
19
14
109
46
57
45
57
120
120
116
Surplus above Similar issuer in Japan

RBS
(a)-(b)
21
12
17
28
38
16
14
29
22
105
44
49
81
8
-11
8
5
-18
72
-73
-2
Shizuoka Bank
Shizuoka Bank
Ito-Yokado
Shoko Chukin
Shoko Chukin
Sony Corp.
Asahi Kasei
Nomura Securities
Japan Airlines
Fuji Heavy Ind.
Fuji Heavy Ind.
Nippon Telecom
NEC Corporation
Yamato Transport
Yamato Transport
Suntory
Suntory
ANA
IHI
Kawasaki Heavy Ind.
Ratings
Aa2/AAAa2/AAAa3/AA+
Aa3/A
Aa3/A
Aa3/A+
A2/ABaa1/BBB+
A3/BB+
Baa3/BB
A- (R&I)
A- (R&I)
AA (JCR)
Baa1/ A3/AA3/ABaa3/ Baa3/ Baa3/BBBaa2/BBBBaa2/BB+
Years
LIBOR
spread
(a)
Rockbottom
spread
(b)
Surplus above
RBS
(a)-(b)
3.9
4.3
3.0
4.1
3.5
2.0
5.0
4.0
5.0
3.0
5.0
5.0
2.0
3.0
3.0
5.0
2.0
5.0
5.0
5.0
4.0
12
13
4
10
7
0
12
23
18
40
30
30
0
8
8
7
10
21
63
21
28
6
6
6
7
7
5
16
43
27
145
149
149
8
39
19
27
140
149
149
75
67
6
7
-2
3
0
-5
-4
-20
-9
-105
-119
-119
-8
-31
-11
-20
-130
-128
-86
-54
-39
* Callable bonds
Source: JPMorgan
Rock-bottom spread analysis for individual bonds

For bonds we looked at in Exhibit 4, we will examine how large surplus each bond has
above the minimum spread required to compensate for credit fundamentals by taking a
market spreads surplus above a rock bottom spread (Exhibit 8).
We can observe that US names have better relative value from credit fundamentals
perspective compared to Japanese names. While most of Japanese names listed in the
exhibit have a large negative surplus, most US names have a large surplus above rockbottom spread. In particular, US single-A names seem to have a relatively large surplus.
Diversification is the key to reduce a risk of a bond portfolio
Finally, we would like to emphasize that buying a portfolio of US corporate bonds
would be better than buying just one or two, because buying a large number of bonds
can reduce a credit risk of a bond portfolio by diversification effects. The US rockbottom spreads in Exhibits 5 and 8 are calculated assuming that the portfolio in which
the bonds will be held is well-diversified. The less diversified the portfolio is, the more
compensation for credit risk is required in buying a corporate bond and the wider a
rock-bottom spread is. Exhibit 9 shows the impact of portfolio diversification on a
rock-bottom spread of a bond that is held in the portfolio.
New York
October 25, 2001


page 96
Exhibit 9
Impact of portfolio diversification on a rock-bottom spread

Five-year 8% coupon Baa2-rated corporate bond
Diversity
score
Rock-bottom
spread
60
59
50
61
40
63
30
66
20
72
10
84
102
176
Source: JPMorgan
New York
October 25, 2001


page 97
Appendix 1
Including names we already looked at, the exhibit lists spreads of US corporate bonds issued by companies that we
consider to be relatively familiar to Japanese investors with 2-5 years to maturity.
Ratings
Issuer
Industry sector
Aa2/A+
Aa2/A+
Aa2/AA
Aa2/AAAa2/AAAa2/AA
Aa2/AA
Aa2/AA
Aa3/A
Aa3/A
Aa3/A
Aa3/AAAa3/AAAa3/A+
A1/A
A1/A+
A1/A+
A1/AAA1/A+
A1/A+
A2/A
A2/A
A2/AAA2/A
A2/A
A2/A
A2/A
A2/A
A2/A
A3/A
A3/AA3/AA3/AA3/AA3/AA3/AA3/BBB+
Baa1/BBB+
Baa1/BBB+
Baa2/BBB
Baa2/BBB
Baa2/BBB
Baa2/BBB
Baa2/BBB
Baa3/BBBBaa3/BBBBaa3/BBBBaa3/BBBBaa3/BBBBaa3/BBB-

Chevron Corp
CITIGROUP Inc
CITIGROUP Inc
Procter & Gamble Co
Wal-Mart Stores
Wal-Mart Stores
Bank One Corp
Bank One Corp
Bank One Corp
JPMorgan Chase & Co
Sony Corporation
Dow Chemical
Goldman Sachs Group
IBM Corp
Statoil
Unilever Capital Corp
Unilever Capital Corp
AT&T Corp
Boeing Capital Corp
GMAC
GMAC
GMAC
Walt Disney Company
Walt Disney Company
Walt Disney Company
American Home Products
BHP Finance USA Ltd
Hertz Corp
Worldcom Inc.
Sun Microsystems Inc
Time Warner Inc
CSX Corp
FedEx Corp
FedEx Corp
Kellogg Co
Kellogg Co
Bausch & Lomb Inc
Delta Air Lines
Delta Air Lines
Raytheon Co
Raytheon Co
Banks
Banks
Oil&Gas
Financial
Financial
Household
Retail
Retail
Banks
Banks
Banks
Banks
Financial
Home Furnishings
Chemicals
Financial
Computers
Oil&Gas
Household
Household
Airlines
Telecom
Financial
Financial
Financial
Financial
Media
Media
Media
Pharmaceuticals
Mining
Auto Manufacturers
Auto Manufacturers
Auto Manufacturers
Auto Manufacturers
Commercial Services
Telecommunications
Computers
Media
Transportation
Transportation
Transportation
Food
Food
Healthcare-Products
Airlines
Airlines
Aerospace/Defense
Aerospace/Defense
Aerospace/Defense
1. Ratings: Moody's / S&P

2. NC: No call; MW: Make whole (at par)
3. Spreads are indicative bids as of June 14, 2001
Coupon
6.625
7.875
6.625
5.7
6.75
6.6
6.55
5.875
5.625
7.625
6.5
6.75
5.75
6.125
8.625
7.625
5.625
6.875
6.75
6.875
7.155
5.625
7.1
5.875
7.5
6.75
5.125
7.3
6.75
5.875
6.69
7.75
6.9
7.75
7.25
8.25
7.875
7.35
7.75
7.25
6.625
6.875
5.5
6
6.75
6.65
7.7
7.25
7.9
6.3
Maturity
6/15/04
5/16/05
10/1/04
2/6/04
12/1/05
12/15/04
8/10/04
10/15/05
2/17/04
8/1/05
2/1/06
12/1/04
2/25/04
3/4/03
4/1/06
8/17/05
4/12/04
5/1/06
11/1/03
11/1/05
10/15/04
3/15/04
9/27/05
1/22/03
7/15/05
1/15/06
12/15/03
2/8/05
3/30/06
3/15/04
3/1/06
5/27/03
9/1/04
6/15/05
1/18/06
6/1/05
5/15/03
8/15/04
6/15/05
5/1/04
2/12/04
2/15/06
4/1/03
4/1/06
12/15/04
3/15/04
12/15/05
5/15/06
3/1/03
3/15/05
Years
3.0
3.9
3.3
2.6
4.5
3.5
3.1
4.3
2.7
4.1
4.6
3.5
2.7
1.7
4.8
4.2
2.8
4.9
2.4
4.4
3.3
2.7
4.3
1.6
4.1
4.6
2.5
3.6
4.8
2.7
4.7
1.9
3.2
4.0
4.6
4.0
1.9
3.2
4.0
2.9
2.7
4.7
1.8
4.8
3.5
2.7
4.5
4.9
1.7
3.7
Features
NC
NC
MW
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC 1
NC
NC
NC
MW
NC
NC
NC
NC
NC
NC
NC
NC 1
NC
NC
NC
NC
NC
NC
NC
MW
NC
NC
NC 1
NC 1
NC 1
NC 1
MW
NC
NC
NC
NC 1
NC
Trsry
spread
83
83
60
78
90
50
57
67
92
100
108
93
78
70
100
102
68
115
63
85
165
117
83
93
118
135
67
77
95
107
150
117
110
133
153
138
160
165
113
125
132
152
107
127
335
285
295
135
190
185
$LIBOR
spread
-23
32
24
-14
26
9
25
4
-1
44
39
53
-16
22
32
45
-31
42
-17
22
127
20
22
50
62
67
-17
32
-56
11
79
58
76
79
85
86
102
132
59
24
39
82
54
54
278
184
223
62
143
132
LIBOR
spread
-27
22
16
-18
15
0
18
-4
-6
32
27
42
-20
18
20
33
-34
29
-20
12
110
17
14
49
49
53
-18
21
-57
8
61
56
64
67
68
72
95
117
47
17
35
65
50
39
249
167
192
47
135
114
Rockbottom
spread
1
1
56
1
3
2
1
2
2
4
5
4
2
2
6
4
3
7
3
5
5
5
7
5
7
9
6
7
9
13
19
14
13
15
19
15
14
109
32
46
46
57
45
57
137
115
120
120
124
116
New York
October 25, 2001


page 98
Appendix 2
Structure of cross-currency asset swap
Cross-currency asset swaps enable investors to convert dollar-denominated cash flows
of the US domestic corporate bonds into -denominated floating rate cash flows. What
an investor does with a cross-currency asset swap is the same as to borrow US dollars
at LIBOR, buy dollar-denominated paper, and sell all future cash flows in US dollar by $/
forward foreign exchange contracts. Thus, it can remove currency and relative curve
risks and leave only credit risks translated into yen. We explain this mechanism with a
simple example below.
Example 1
Now assume that a yen-based investor buys a five-year dollar-denominated Baa3 corporate bond with 120bp spread over Treasury and asset-swaps it into yen. The swap
transaction that the investor will enter here can be divided into two parts: (1) Asset swap
transaction that converts a dollar-fixed rate into a dollar-floating rate and (2) Crosscurrency swap transaction that converts dollar-denominated floating cash flows into
yen-denominated cash flows.
Principal cash flows
A cross-currency swap converts the $10 million dollar principal into 1.2 billion of yen
(assume that $/ exchange rate is 120).
JPMorgan
1.2 billion
Investor
$10 million
$10 million
1.2 billion
$10 million
5-year Baa3
corporate bond
Coupon 7.25%
$10 million
$10 million
Cash flow at value date
Cash flow at maturity
Interest cash flow

Through a cross-currency asset swap, T+120bp fixed rate flows are swapped into
$LIBOR+45bp floating rate flows, and then into LIBOR+30bp floating rate flows. The
15bp difference between a $LIBOR spread and a LIBOR spread represents the difference of yield curves in the US and Japan.
T+120bp
Investor
JPMorgan
$L+45bp*
$L+45bp*
T+120bp
5-year Baa3
corporate bond
Coupon 7.25%
$10 million
L+30bp
* US dollar LIBOR cash flows between MGT and an investor do not occur as the two opposite flows will be
offset with each other.
New York
October 25, 2001


page 99
Example 2
If an investor wants to avoid separately managing swap cash flows and bond cash
flows, it is possible to repackage the above cash flows by using an SPC as illustrated
below. Please note that an all-in spread that investor receives will be narrower compared
to Example 1 due to a cost involved in establishing an SPC.
Principal cash flows
Repackaged bond issued by an SPC
1.2 billion
Investor
5-year Baa3
corporate bond
Coupon 7.25%
$10 million
1.2 billion
Cash flow at value date

$10 million
Cash flow at maturity
1.2 billion
$10 million
1.2 billion
JPMorgan
Interest cash flows

Repackaged bond issued by an SPC
Investor
5-year Baa3
corporate bond
Coupon 7.25%
$10 million
L+30bp
- SPC cost
$L+45bp
T+120bp
L+30bp
$L+45bp
JPMorgan
The structure explained here might be exposed to an foreign exchange risk if the
underlying security (5-year Baa3 US corporate bond in this case) defaults. If the
underlying bond defaults, the currency swap position will over-hedge the residual
principal (recovery value) and the remaining coupons. An investor might take an
foreign exchange loss when it tries to unwind the over-hedging swap position depending on the market condition at the time of unwinding. To avoid taking such a risk, one
can build the same structure by using a swap that can be cancelled when the underlying
bond defaults. (Using a cancelable swap will take an extra cost compared to a normal
swap.)
New York
October 25, 2001

Portfolio Research

page 100
For further information on Rock-bottom Spreads

contact:
New York
Peter Rappoport
Frank Zheng
London
Luca Brusadelli
Tokyo
Tatsushi Kishimoto
(1-212) 834-7046
(1-212) 834-9224
(1-212) 834-9226
(44-20) 7325-5607
(81-3) 5573-1521
Additional information is available upon request. Information herein is believed to be reliable but JPMorgan does not warrant its completeness or accuracy. Opinions and estimates constitute our judgment and are subject to change
without notice. Past performance is not indicative of future results. The investments and strategies discussed here may not be suitable for all investors; if you have any doubts you should consult your investment advisor. The investments
discussed may fluctuate in price or value. Changes in rates of exchange may have an adverse effect on the value of investments. This material is not intended as an offer or solicitation for the purchase or sale of any financial instrument.
JPMorgan and/or its affiliates and employees may hold a position or act as market maker in the financial instruments of any issuer discussed herein or act as underwriter, placement agent, advisor or lender to such issuer. Copyright
2001 J.P. Morgan Chase & Co. All rights reserved. JPMorgan is the marketing name for J.P. Morgan Chase & Co., and its subsidiaries and affiliates worldwide. J.P. Morgan Securities Inc. (JPMSI), member of NYSE and SIPC. Morgan
Guaranty Trust Company of New York and the Chase Manhattan Bank, members of FDIC. J.P. Morgan Securities Asia Pte Ltd., (JPMSA) and Chase Manhattan Asia Ltd., are regulated by the Hong Kong Securities & Futures Commission.
JPMSA is regulated by the Monetary Authority of Singapore and the Financial Services Agency in Japan. J.P. Morgan Futures Inc., is a member of the NFA. Issued and approved for distribution in the UK and the European Economic
Area by J.P. Morgan Securities Ltd., Chase Manhattan International Limited and J.P. Morgan plc, members of the London Stock Exchange and regulated by the FSA. Issued and distributed in Australia by Chase Securities Australia
Limited and J.P. Morgan Australia Securities Limited which accept responsibility for its contents and are regulated by the Australian Securities and Investments Commission. J.P. Morgan Markets Australia Pty Ltd. is a licensed investment
adviser and futures broker member of the Sydney Futures Exchange. Clients should contact analysts at and execute transactions through a JPMorgan entity in their home jurisdiction unless governing law permits otherwise.

No1 Rock-Bottom Spreads

Uploaded by

Copyright:

Available Formats

No1 Rock-Bottom Spreads

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

No1 Rock-Bottom Spreads

Uploaded by

Copyright:

Available Formats

New York

October 25, 2001

J.P. Morgan Securities Inc.

The cashflow pattern, maturity and seniority of the bond;

Rock-Bottom Valuation: the basic recipe

Do not buy the

Buy if surplus spread

INVESTMENT STRATEGIES: NO. 1

J.P. Morgan Securities Inc.

So, rock-bottom spreads allow you to remove credit

Rock-Bottom versus Market Spreads for US Corporate Bonds

Investment Strategies: No. 1

Anatomy of rock-bottom spread differences

and with historical

and with historical

and with current HY

and with current HY

B-rated corporates have rarely paid in excess of their

those with lower rock-bottom spreads than market,

Investment Strategies: No. 1

1. Understanding Rock-Bottom Spreads

Annualized return relative to Index

Another resource for understanding rock-bottom

2. The Rock-Bottom Spread Web Calculator

US High Yield Rock-Bottom Bondpicking Strategy

One of the principal drivers of rock-bottom spreads is

US Investment Grade Rock-Bottom Bondpicking Strategy

J.P. Morgan Securities Inc.

1. Understanding Rock-Bottom Spreads

Rock-Bottom Spread Mechanics

Valuing European Credit Fundamentals

2. The Rock-Bottom Spread Web Calculator

Introducing the Rock-Bottom Roundup

3. Rock-Bottom Investment Strategies

Picking Investment Grade Bonds

Valuing Rating-triggered Step-up Bonds

Comparing Credit Fundamentals:

Emerging Markets versus High Yield:

US Credits Look Attractive for Japanese Investors

J.P. Morgan Securities Inc.

amount of data involved. This problem is solved by the

3. Rock-Bottom Investment Strategies

Investment Strategies: No. 1

One of the virtues of the rock-bottom framework is that

J.P. Morgan Securities Inc.

Investment Strategies: No. 1

Valuing Credit Fundamentals:

Rock-Bottom Spreads vs. Market Spreads,

Rock-bottom spreads calculated from the credit downswing scenario

market spreads. Speculative grade market spreads are

J.P. Morgan Securities Inc.

Adverse Credit Outlook: Change in Rock-Bottom Spreads

Investment Strategies: No. 1

6. Interpreting Market Spreads ....................................... p.8

A rock-bottom spread calculator is available on the Credit

for high-yield investors, and is made easier by tracking rockbottom spreads.

J.P. Morgan Securities Inc.

Investment Strategies: No. 1

Components of Credit Fundamentals

Credit Fundamentals: Historical Average Scenario